Mike Hammett

How things actually work.

Tag: Redundancy

  • Federal Rules Let 911 Run Without Diverse Paths

    Federal Rules Let 911 Run Without Diverse Paths

    Mike Hammett has spent more than 20 years building and operating ISP, voice, internet exchange, and colocation networks.

    This article is part of a series on government network outages. The other pieces will be published as they are completed.

    TL;DR

    • By my count, at least 29 network failures since 2009 have taken down 911 or local alerting, 25 of them with a detailed published cause. Four of those 25 began with an error during maintenance or a network change, which can break any network, though in two of them a design gap let the error spread. In each of the other 21, one failure alone took service down, or two failures shared a cause, or a second failure came before the first was repaired.
    • They include a Colorado routing facility that stopped sending calls to 81 call centers in seven states [1], four supposedly diverse paths that met in one Houston facility, a “temporary” repair left in place for 11 years [2], and one AT&T switch that 73 of Oklahoma’s 122 call centers were connected to [3][4].
    • In June 2026, the FCC voted to replace the annual 911 diversity certification with a one-time filing [5]. After proposing to require that providers eliminate single points of failure, it adopted “reasonably sufficient mitigation” of them instead [5].
    • State 911 authorities can contract for what the federal rule does not require. California’s new plan keeps a statewide backup provider behind its primary one [6], and Nebraska’s new solicitation demands no single point of failure [7].

    911 Is a Government Service on Someone Else’s Network

    911 is a government service that runs mostly on networks the government does not own. State and local 911 authorities contract with carriers and vendors for routing, transport, and call delivery, and commercial carriers carry most calls to the call centers [8]. When those networks fail, the public cannot reach help. A carrier answers to its customers. A 911 authority answers to everyone who will ever need to call, so its contracts have to demand more than a carrier would build on its own.

    The Standard on Paper

    The FCC adopted its 911 reliability rule in December 2013, after the June 2012 derecho, which its public safety bureau found had disrupted 911 mostly through preventable failures, among them critical circuits that lacked physical diversity [9]. Under that rule, covered carriers certified, under penalty of perjury, that they had audited critical 911 circuits for physical diversity and tagged them so later rearrangements would not quietly collapse that diversity [9][10]. A carrier that could not certify an element could instead show alternative measures “reasonably sufficient to mitigate the risk of failure” [10]. The rule states the physical standard plainly: “Circuits or paths that share a common segment such as a fiber-optic cable or circuit board are not physically diverse even if they are logically diverse for purposes of transmitting data” [10].

    For legacy circuits, the covered path runs from the selective router to the central office that serves each call center, and the FCC’s guidance says the diversity audits apply to the facilities between those two points [10][11]. The same guidance says the “local loop” between each call center and its serving central office is not subject to the audits [11]. That last segment is where one cable or one building entrance can undo everything upstream. In 2009, the FCC’s public safety bureau had written that “in the ideal situation, diversity should be gained by connection to separate central offices,” and that separate entrance facilities at the call center were also desirable [12].

    In June 2026, the FCC voted to replace the annual certification with a one-time certification, updated only after material changes [5]. Providers that filed annual certifications under the 2013 rule file none during the transition, and the first one-time certification is due 18 months after the FCC’s public safety bureau announces that the Office of Management and Budget has approved the new filing requirements [5]. It had proposed requiring providers to “eliminate all single points of failure rather than mitigate them,” and adopted a benchmark of “reasonably sufficient mitigation of single points of failure” instead [5]. For IP paths, the rule now lets a provider meet the physical diversity benchmark “by implementing automatic rerouting capabilities, load balancing, and geographically-distributed routing facilities, transport nodes, and node links sufficient to reasonably mitigate the risks of single points of failure,” and it counts VPN service over the public internet among the “automatically re-routed paths” [10]. The same order brings major IP transport and IP 911 aggregation providers under the rules for the first time, and it lets a statewide, territorial, or tribal 911 authority request a provider’s certifications and reports directly, under a confidentiality agreement [5].

    The 911 Record

    On April 9, 2014, a 911 routing facility run by Intrado in Englewood, Colorado, stopped sending calls to 81 call centers in seven states [1]. A software counter that assigned an identifier to each call stopped counting at 40 million [13]. The routing system logged the failure as low level, a default setting Intrado had accepted, and raised no major or critical alarm [1]. More than 11 million people lost 911 service for up to six hours, until Intrado switched the traffic to its Miami hub, which restored service almost at once [1]. The FCC fined CenturyLink $16 million and Intrado $1.4 million [14], and its report flagged the trend of carriers and call centers contracting 911 functions out to third parties [1].

    On December 27, 2018, a switching module in a CenturyLink optical node in Denver generated four malformed management packets [15]. They traveled over a proprietary management channel that the equipment’s maker, Infinera, ships enabled by default and that CenturyLink knew about but neither configured nor used, and the nodes kept retransmitting them to one another [15]. The outage lasted almost 37 hours and affected as many as 22 million customers in 39 states, including about 17 million in 29 states who lacked reliable access to 911 [15]. The FCC found that “at least 886 calls to 911 were not delivered” [15]. Washington’s utility commission says the outage left residents there without 911 for almost 50 hours [16].

    On June 15, 2020, a fiber transport link in T-Mobile’s network failed for about 12 minutes, and the outage that followed lasted more than 12 hours [17]. T-Mobile had misconfigured the weight of the links to one of its routers, so traffic from the failed link went to a router that “was not provisioned to process the signaling traffic that the failed link carried,” and a software flaw had been latent for months [17]. The outage kept 23,621 of the 134,874 calls to 911 that reached T-Mobile’s network from reaching call centers [17].

    On July 11, 2025, Pennsylvania’s statewide NG911 system had intermittent outages, and the state warned the public by Wireless Emergency Alert [18]. A preliminary report traced them to an operating system defect in a core that runs four instances in two data centers [19].

    Local governments depend on the same commercial networks. During the August 8, 2023, Lahaina wildfire, all 21 cell towers serving West Maui experienced what Maui County calls “total cell service failure” while the county sent at least 14 Wireless Emergency Alert messages telling residents to evacuate [20]. The county says it did not know about the failures during its response. In 2024, it sued the wireless carriers, alleging they did not give its 911 operators the outage notice federal rules require within 30 minutes [20]. The FCC reported some 911 calls were not going through because the Lahaina switch was isolated [21]. Fiber strung on power poles burned, while underground fiber through Lahaina survived [22]. Its owner, Sandwich Isles Communications, had buried it in 2009, and its founder, Albert Hee, offered the surviving strands to the carriers [22].

    Nashville 2020: One Building, 66 Districts

    On Christmas morning 2020, a bomb in an RV exploded outside AT&T’s central office at 185 Second Avenue North in Nashville. The blast knocked out the building’s primary power, and flooding took out its emergency backup generators [23].

    The 911 damage was wider and longer. Sixty-six Tennessee emergency communications districts had service interruptions lasting more than 97 hours [24]. Some 911 centers did not get AT&T’s outage notices for six days, in some cases because the link that carried them had been damaged [25]. Nashville’s own emergency communications center kept outbound calling by using older phones from a different carrier, pre-positioned at each console, and APCO’s account of the event concluded that carrier diversity is “an absolute must” [26]. One building held the switching and the power for a region, and the path that told 911 centers what was happening broke along with them.

    Failures Far From Home

    Aggregation puts many services behind one point, and the outage lands wherever the services are, regardless of where the failure is. In December 2016, a fire at a Sprint building in Omaha left Sprint customers in the Twin Cities unable to reach 911 [27][28]. On August 1, 2018, Minnesota’s statewide 911 failed for about an hour after a maintenance error by a technician at its vendor, West, disrupted the vendor’s Miami hub [29]. On January 12, 2021, a bad card supporting a large national fiber in Green Bay, Wisconsin, left dispatchers in eight southeastern Minnesota counties unable to hear 911 callers for 7 hours and 12 minutes [30]. In April 2024, a fiber cut in Kansas City took 911 down across South Dakota and in parts of Nebraska and Nevada [31].

    In January 2026, a cable cut left the 911 center in Park County, Wyoming, with a single cell phone as the public’s only way to reach it for about three hours, and service was not fully restored for days [32][33]. Monte McClain, Park County’s communications supervisor and 911 manager, said the state’s 911 calls all route through one selective router in Cheyenne, and if the line between the county and Cheyenne is cut, “we go dead” [32]. On September 23, 2026, a switch AT&T uses to route 911 calls in Oklahoma failed. The state’s 911 coordinator said 73 of its 122 call centers were connected to that switch, and for about seven hours their calls went to the wrong centers or to backup ten-digit numbers [3][4].

    The FCC’s 2014 report on the Colorado failure warned that “a 911 outage in an IP-supported network has the potential to affect a much greater number of PSAPs and people, across multiple states” [1]. The Nebraska commission told its State 911 Department that reaching five nines may require eliminating aggregation points [2]. Aggregation is cheaper, with fewer routing centers, fewer circuits, and one contract, and each consolidation widens the area one failure reaches.

    Before California’s next-generation design went in, a network examination for the state’s Public Utilities Commission found that only 41 of the 93 call centers hosted at Frontier central offices had confirmed diverse connections [34]. Cal OES split the state into four regions, each served by its own regional network provider, and added a prime provider to give every call center a second level of service over “separate unique paths to each PSAP” [35]. The national 911 program called the design “highly effective for ensuring physical and logical diversity is provided from call inception to termination” [35]. Between 2021 and 2024, only 23 of the state’s roughly 450 dispatch centers began moving voice calls onto the regional networks, and some “reported routing problems, outages, and dropped calls” [6]. The state’s Legislative Analyst’s Office reports that “complexities of the network and interdependencies between the service providers created failure points” [6]. In November 2025, Cal OES switched to a plan with one primary statewide provider and one backup provider, and it now expects to retire legacy 911 in 2030 [6]. The failures in this section come from aggregation onto one provider and one set of paths. California’s came from providers that depended on each other. A properly designed network keeps the reach of a single failure no larger than the place where it happened. That means regional routing that can stand alone, local fallback when the national core is unreachable, and paths that leave each region by different routes.

    Nebraska’s Audit Stopped at the Incumbent’s Plant

    Nebraska’s Public Service Commission investigated three Lumen 911 outages and issued its findings in January 2025 [2]. On August 31, 2023, two separate fiber cuts, one in Minnesota and one in Omaha, left Nebraskans unable to reach 911 for over 10 hours [2]. The Minnesota cut came first and stayed unrepaired until 5:31 a.m. on September 1 [2]. The commission found “the lack of diversity and redundancy was the root cause of this failure” [2]. On April 17, 2024, a single cut in Kansas City “revealed a single point of failure” tied to “a repair that was put in place about a decade ago which was supposed to be temporary” [2]. The ring had been collapsed for 11 years, and Lumen had “not completed a full physical fiber audit in at least 11 years” [2]. On July 9, 2024, four paths that were supposed to be diverse “all four met in the same facility in Houston,” and a power loss there cut off 14 Nebraska call centers [2]. It also found that Lumen’s 911 system “at best performed at three nines and not the five nines guaranteed by Lumen” [2]. A guarantee the buyer cannot check protects the invoice and little else.

    There is usually far more fiber, and more diverse fiber, than a carrier’s diversity audit admits. A real audit looks past the incumbent’s own plant to the local cable company, the local telephone company, highway department fiber, regional and national fiber networks, local ISPs, and the electric utility. It also counts the middle-mile transport that networks run through other networks’ territory on the way between towns they serve. Near where I live in northern Illinois, Comcast fiber crosses a town Mediacom serves on its way from one Comcast town to another. Nebraska law requires its next-generation 911 contractor to report each year on “the capabilities and redundancies of the next-generation 911 service network” [36]. The 2024 diversity audit filed under that law, while Lumen held the contract, lists 13 call centers that were not fully diverse, and the reasons and alternative measures it gives concern the provider’s own cables, cross-connects, common equipment, and routes to the incumbent carrier [37][38].

    The Nebraska Public Power District had installed 1,115 miles of fiber by 2011, in and around South Sioux City, Norfolk, Lincoln, North Platte, Grand Island, Kearney, and other towns, and the Public Service Commission has approved the prices at which it leases dark fiber to carriers [39]. The Omaha Public Power District maintains 621 miles of optical ground wire on its power lines and owns 200 miles of underground fiber [40]. The state Department of Transportation lists 144-strand fiber along I-80, N-370, N-50, and US-75 in Sarpy County and the Omaha metro [41], and Sarpy County is home to Offutt Air Force Base, whose call center is one of the 13 in the audit [37].

    Hamilton Telecommunications, the local exchange carrier in Hamilton County, completed its first fiber SONET ring in 1985 and, in 2010, joined six other local telecommunications companies to form NebraskaLink [42], which won a federal grant toward a middle-mile network of nearly 2,000 miles across nearly the entire state [43].

    In cloud engineers’ shorthand, pets are unique systems kept alive by hand, and cattle are interchangeable units in arrays built for failure [44]. Many incumbent telephone companies treat their infrastructure as pets instead of cattle. They build everything in a town around one poorly maintained building in the middle of the wire center and the plant around it, and when diversity comes up, that building is the only facility they think of. A second route does not have to end there, and no one company has to own all of it. Fiber from the electric utility, the highway department, a regional network, and a local ISP can be chained into a path that lands at a different central office, a different data center, or the other node of the 911 core. Next-generation 911 has moved from switched-access voice to IP packets, and packets can reach the core over any provider’s path. The FCC’s June 2026 order anticipates that providers may use “general Internet paths as diverse pathways” for 911 traffic [5].

    Nebraska’s audit covers the path from the IP core to each call center, and the reasons it gives for missing diversity include channel banks, digital cross-connect systems, and SONET, all gear from the circuit-switched telephone network [37]. Eight of the 13 entries cite central office avoidance as a reason the circuits are not fully diverse, and six say full diversity was economically infeasible [37]. For the Hamilton County Sheriff, the entry reads “Only one fiber route to ILEC” and says a second would take more than 15 miles of new fiber and the incumbent’s agreement [37]. Pierce County’s entry describes the same kind of build, and no entry proposes a route over another owner’s fiber or to a different building or node [37]. Those two entries and six others list “Circuits are logically diverse” as an alternative measure, which the federal rule does not count as physical diversity [10][37]. The audit also asks the provider to list every call center with physically diverse circuits. The only entry is “Whitefish County Flathead Backup,” and the audit does not say where that is [37]. Whitefish is in Flathead County, Montana, where the county’s 911 center had a backup system in 2017 [45]. Every location’s cost is different, but a second path rarely means building a new network from scratch. The expensive pieces are likely already there, just owned by somebody else.

    In July 2026, Nebraska posted its intent to award the statewide next-generation 911 network and core services contract to AT&T, over four other bidders, including Lumen [46][47]. The solicitation requires the network to “ensure no single point of failure through diverse network paths for the entire system” [7]. Lumen’s bid to keep the contract said that where it could not provide a diverse path to a call center, it recommended “a common access provider to diverse network POP locations” and “a tertiary over-the-air solution,” such as low Earth orbit satellite or 5G/LTE [48]. With a common access provider, both circuits can still share one trench into the building, however many POPs they reach.

    The solicitation also requires core data centers “separated by at least 100 miles,” network diagrams that show “physical and logical path diversity,” out-of-band management and monitoring traffic, failover testing before a change is accepted, and proactive tracing of IP routing within the diverse and redundant paths [7]. Bidders had to submit a diversity plan covering route and provider diversity, facility and region separation, and power, and show how a fault in a provider’s network for another state would be kept out of Nebraska’s [7]. The winner must also identify the places with limited physical diversity or single-threaded resources and propose a mitigation for each [7]. Its maintenance procedures must not create single points of failure during changes [7]. That is the standard the FCC declined to adopt, written into a solicitation by a state that lived through the failures. It holds only if the state checks the diversity plan against route records down to the fiber, because the physical layer is where Nebraska’s own audit found its gaps.

    The Same Chains of Contractors

    911 networks run through chains of contractors, and each link sees only its own piece. In December 2018, TeleCommunication Systems, which provided next-generation 911 service in Washington and north central Texas, stopped receiving 911 calls from those areas for stretches because its signaling links ran over CenturyLink’s network, and its signaling provider said it could not provide any alternate routing [15]. The FCC’s new order lets a provider that relies on third-party facilities certify to “network practices within their control” and cite the representations third parties make in service level agreements [5]. Building a route from several companies is sound practice, but a rule that lets each company vouch only for its own piece leaves no one certifying the segments they share.

    The FCC’s Case for Mitigation

    The FCC’s case is that IP networks are built differently. Providers told it that eliminating all single points of failure would be cost prohibitive or infeasible. Next-generation 911 networks route calls dynamically, they said, so packets can converge at single points without the provider knowing. The FCC agreed that “eliminating all single points of failure is not presently a design goal of typical NG911 deployments” [5]. It called auditing and tagging legacy TDM practices that “do not necessarily measure the inherently more-resilient geodiversity of IP networks” [5]. It also left the choice to states. A state 911 authority, it wrote, may decide that “paying for diverse IP transmission paths to remote or rural PSAPs is cost prohibitive” and use policy routing to send calls to other call centers when one goes offline [5]. It ended the annual certification, concluding that “annual filing is not an effective means of obtaining timely and actionable information” and that it burdens providers whose networks have not changed [5]. Under the new rule, providers file an update within 90 days of finding a material change, and they judge for themselves what counts as material [5]. For legacy circuits, the rule still names yearly diversity audits as the way to meet the standard, but providers no longer certify them each year [5][10].

    That measures the rule against how NG911 networks are built today, and the record above shows what that design produces. The FCC’s own release announcing the June 2026 order said it had seen “a corresponding increase in major, multi-state 911 service outages” as the nation moved to NG911 over the past decade [49]. The need for diversity does not go away when 911 moves from TDM to IP. Packets still ride fiber, and two IP paths that share a trench, a conduit, or a building entrance fail together, as two TDM circuits would.

    Geodiversity in the core does nothing for a call center with one fiber route to the incumbent’s office. Rerouting calls to a neighbor is a real tool, and it works only when the neighbor’s paths do not run through the same failure. In January 2026, the cable cut that isolated Park County’s 911 center in Wyoming also disrupted 911 calls in neighboring Big Horn County, and both counties’ calls were eventually sent to dispatchers in Worland [33].

    The point of an annual audit is to find a gap before it takes 911 down, and that is worth the burden. A filing updated only when a provider finds a change leaves the finding to the next outage. Lumen’s Kansas City gap was found by a fiber cut, after at least 11 years without a full physical fiber audit [2]. The failures under the old rule mean the FCC needs to get tougher with providers, not weaker.

    What Would Change the Record

    The first change is the standard the FCC proposed and then dropped. Providers should have to eliminate single points of failure on 911 paths, audit them every year under an accountable official’s certification, and carry the audit past the serving central office to the call center’s own building entrance.

    The second is enforcement. The federal rule requiring carriers to notify potentially affected 911 call centers of outages within 30 minutes [50] exists for counties like Maui, which says it sent evacuation alerts to towers it did not know were down [20]. The FCC should enforce it against every carrier that leaves a county sending alerts to dead towers.

    The third is disclosure. Carriers already report outages to the FCC on a 72-hour and 30-day clock [50]. Every 911 outage that blocks calls should end with a published reason for outage on the same clock. Nebraska’s solicitation requires a root cause analysis for any system trouble, but it lets the contractor mark those analyses, and its detailed network maps, as confidential [7]. The Nebraska commission’s order shows what a published report can look like. It named the 11-year temporary repair, the collapsed ring, and the four paths that met in Houston [2].

    The fourth is the contract, which state 911 authorities control now. Every 911 contract can require diverse paths into each call center, separate power behind them, monitoring of the backup path, and out-of-band management of the provider’s critical network elements. It can also require failure domains that keep one region’s failure out of the others and are tested on a schedule with results reported to the state, and repair times on the redundant half, with remedies beyond credits.

    The diversity audit has to count every owner’s fiber, end the second path at a different central office, data center, or core node, and map any chained route to the trench again after every change. A state without its own network engineer needs one, on staff or independent of every bidder, to write the requirements, read the route maps, and witness the tests. It can also use its new right to ask providers for their reliability information.

    The last is a third path. For the largest call centers and the sites that run the 911 core, two wireline paths are not enough. A third path that is not wireline at all, such as licensed microwave or satellite, cannot share a trench, a conduit, or a central office with the other two. It only has to carry 911 calls and their location data, but it has to be independent in every other way, with its own building entrance, power, carrier, and management. Lumen’s own Nebraska bid recommended an over-the-air third path where it could not provide a diverse one [48], and Indiana’s 911 board “has implemented tertiary connections from third party service providers” where legacy circuits proved unreliable [51]. The test for the third path is whether it fails for reasons the other two do not share. A cellular path can ride the same fiber and lose power in the same storm as the wireline paths it is meant to back up. In West Maui, burned overhead fiber took out 911 and cell service together [22]. Using one satellite network for every call center’s third path passes that test, because an outage of that network leaves the two wireline paths up. Starlink’s cheapest business plan, for example, costs $55 a month plus equipment [52][53].

    Quick Compare: 911 Outages

    The table counts network failures since 2009 that this article found documented and that took down 911 or local alerting, even briefly. It is not a complete list. Four rows are marked Unknown because this article found no published account that shows what was single, and they stay in the count. The pattern claim in the TL;DR covers the other 25, except the four that began with an error during maintenance or a network change. The last column is this article’s reading of each cited cause.

    DateServiceReported causeWhat was single
    April 9, 2009911 in Santa Clara, Santa Cruz, and San Benito countiesFiber cut in manholes by unknown attackers [54][55]Unprotected fiber route
    October 29 and 30, 2012New York area 911Sandy flooded central offices and their backup power [56]Backup power in a flood zone
    April 16, 2013Some 911 systems south of San JoseFiber cut in a vault before the substation attack [57][58][59]Unprotected vault
    April 9 and 10, 201481 call centers in seven statesRouting counter hit its limit; alarms rated low-level [1][13]One routing center
    December 13, 2016Sprint 911 in Omaha and the Twin CitiesElectrical fire at Sprint’s Omaha exchange [27][28]One building
    March 8, 2017AT&T wireless 911, about 12,600 callersNetwork change pushed a bad whitelist record to the live 911 network [60]Change error; one logical path for two vendors
    August 1, 2018Minnesota statewide 911Maintenance error at the Miami hub [29]Maintenance error; backup hub partly failed
    December 27 to 29, 2018911 in 29 statesMalformed packets on an unused management channel enabled by default [15]One transport network
    June 15, 2020T-Mobile 911 nationwideTraffic failed over to a router not provisioned for it [17]Backup on paper
    September 28, 2020911 in ten statesIntrado traffic manager work; Lumen failed with it [61]Change error, propagated
    December 25, 2020911 in 66 Tennessee districtsBombing cut central office power; flooding took the generators [23][24]One central office
    January 12, 2021Eight southeastern Minnesota countiesBad card on national fiber in Green Bay [30]Remote aggregation point
    August 8 and 9, 2023West Maui 911 and county alertsFire burned overhead fiber; all 21 towers failed [20][21][22]Overhead fiber route
    August 31 to September 1, 2023Nebraska 911, 41 call centersTwo fiber cuts, the first unrepaired [2]Unrepaired first failure
    September 2 and 3, 2023Four Nebraska countiesFire at Windstream’s Lincoln facility [62][63]One facility
    December 12, 2023Six Indiana counties and the Indianapolis airportDisruption to servers hosted by AT&T, per the state [64]One provider’s host servers
    February 22, 2024AT&T wireless, more than 25,000 calls to 911Misconfigured network element put into production [65]Maintenance error
    April 17, 2024Nebraska, South Dakota, NevadaFiber cut on an 11-year temporary repair [2][31]Collapsed ring
    May 30, 2024Landline 911 in parts of IowaPipe break forced a power-down in a Lumen building [66][67]One building’s power
    June 18, 2024Massachusetts statewide 911Vendor firewall blocked calls [68]One network element
    July 9, 2024Nebraska 911, 14 call centersPower loss in Houston [2]Four paths, one facility
    May 29, 2025Five southeast Nebraska countiesFiber cut on Lumen’s network [69]Unknown
    July 11, 2025Pennsylvania statewide NG911Operating system defect in the 911 core [18][19]One software defect
    September 25, 2025Parts of Mississippi, Louisiana, AlabamaThird-party fiber cuts in AT&T’s network [70]Unknown
    January 5 to 9, 2026Park and Big Horn counties, WyomingFiber cut on the path to the state’s one selective router [32][33]One path, one router
    January 14, 2026Verizon wireless nationwide, including some calls to 911Software issue, per Verizon [71][72][73]Unknown
    June 15, 2026Parts of four statesCause not published [74][75]Unknown
    August 6, 2026Dakota County 911511 Building outage, likely a power failure; backup power also affected [76][77]One building
    September 23, 2026Oklahoma 911, 73 of 122 call centersOne AT&T routing switch failed [3][4]One switch

    Cover photo: USDAgov, U.S. Department of Agriculture / Wikimedia Commons, public domain

    Sources

    [1] FCC Public Safety and Homeland Security Bureau, April 2014 Multistate 911 Outage: Cause and Impact, Report and Recommendations, PS Docket No. 14-72, October 2014. https://docs.fcc.gov/public/attachments/DOC-330012A1.pdf

    [2] Nebraska Public Service Commission, Order Issuing Findings and Closing Investigation, Application Nos. 911-075/PI-248 and 911-077/C-5581/PI-252, January 14, 2025. https://nebraska.gov/psc/orders/state911/2025-01-14%20911-075%20PI-248%20911-077%20C-5581%20PI-252%20Order%20Issuing%20Findings%20and%20Closing%20Investigation.pdf

    [3] Lauren Henry, KFOR, via Yahoo News, “Disruption at 911 call centers caused by switch failure,” September 24, 2026. https://www.yahoo.com/news/us/articles/disruption-911-call-centers-caused-230502600.html

    [4] Olivia Hickey, KOCO, “911 outage disrupts emergency calls across Oklahoma for hours,” September 24, 2026. https://www.koco.com/article/911-outage-disrupts-emergency-calls-across-oklahoma/73876198

    [5] FCC, Facilitating Implementation of Next Generation 911 Services (NG911); Improving 911 Reliability, Second Report and Order and Second Further Notice of Proposed Rulemaking, FCC 26-39, adopted June 25, 2026, released June 26, 2026. https://docs.fcc.gov/public/attachments/FCC-26-39A1.pdf

    [6] Legislative Analyst’s Office, “The 2026-27 Budget: Office of Emergency Services Next Generation 911 System,” February 27, 2026. https://lao.ca.gov/Publications/Report/5145

    [7] Nebraska Public Service Commission, Request for Proposal 202691101, Emergency Services IP Network (ESInet) and Next Generation Core Services, February 10, 2026. https://das.nebraska.gov/materiel/purchasing/RFP%20202691101/2026-2-10%20911%20RFP%20FINAL.pdf

    [8] FCC, Sixteenth Annual Report to Congress on State Collection and Distribution of 911 and Enhanced 911 Fees and Charges, for calendar year 2023, December 30, 2024. https://www.fcc.gov/sites/default/files/16th-Annual-911FeeReport-2024.pdf

    [9] FCC, Improving 911 Reliability, Report and Order, FCC 13-158, December 2013. https://docs.fcc.gov/public/attachments/FCC-13-158A1_Rcd.pdf

    [10] 47 CFR § 9.19 (Legal Information Institute). https://www.law.cornell.edu/cfr/text/47/9.19

    [11] FCC, Frequently Asked Questions: FCC 911 Reliability Certification, updated February 16, 2024. https://www.fcc.gov/frequently-asked-questions-fcc-911-reliability-certification

    [12] FCC Public Safety and Homeland Security Bureau, Tech Topic #14, route diversity for PSAPs (archived May 8, 2009). https://web.archive.org/web/20090508103035/http://www.fcc.gov/pshs/techtopics/techtopics14.html

    [13] Robert N. Charette, IEEE Spectrum, “FCC Chairman Calls April’s Seven State Sunny Day 911 Outage ‘Terrifying’,” October 27, 2014. https://spectrum.ieee.org/-fcc-chairman-calls-aprils-seven-state-sunny-day-911-outage-terrifying-

    [14] FCC, “FCC Fines CenturyLink and Intrado $17.4 Million for Multi-State 911 Outage,” April 6, 2015. https://docs.fcc.gov/public/attachments/DOC-332853A1.pdf

    [15] FCC Public Safety and Homeland Security Bureau, December 27, 2018 CenturyLink Network Outage Report, August 19, 2019. https://docs.fcc.gov/public/attachments/DOC-359134A1.pdf

    [16] Washington Utilities and Transportation Commission, “CenturyLink faces $7.2 million penalty for 2018 911 outage,” Docket UT-181051, 2020. https://utc.wa.gov/news/2020/centurylink-faces-72-million-penalty-2018-911-outage

    [17] FCC Public Safety and Homeland Security Bureau, June 15, 2020 T-Mobile Network Outage Report, PS Docket No. 20-183, October 22, 2020. https://docs.fcc.gov/public/attachments/DOC-367699A1.pdf

    [18] Pennsylvania Emergency Management Agency, “Shapiro Administration Announces Statewide 911 Outage Has Been Resolved,” July 2025. https://www.pa.gov/agencies/pema/newsroom/2025/shapiro-administration-announces-statewide-911-outage-has-been-r

    [19] Central Bucks News, “9-1-1 outage caused by operating system ‘defect,’ not cyberattack,” July 24, 2025. https://centralbucksnews.com/news/2025/jul/24/9-1-1-outage-caused-by-operating-system-defect-not-cyberattack

    [20] Peter Boylan, Honolulu Star-Advertiser, via Hawaii Tribune-Herald, “Maui County sues cell carriers for noncompliance amid fires,” May 3, 2024. https://hawaiitribune-herald.com/?p=320924

    [21] RCR Wireless, “FCC: 95% of cell sites out in areas affected by Maui’s wildfires,” August 14, 2023. https://www.rcrwireless.com/20230814/featured/fcc-95-of-cell-sites-out-in-areas-affected-by-mauis-wildfires

    [22] Kevin Dayton, Honolulu Civil Beat, “Fragile Fiber Lines In Lahaina Were Burned And Battered, Taking Out 911 And Cell Service. Now What?” August 18, 2023. https://civilbeat.org/2023/08/fragile-fiber-lines-in-lahaina-were-burned-and-battered-taking-out-911-and-cell-service-now-what/

    [23] Rich Miller, Data Center Frontier, “Nashville Explosion: AT&T Repairs Damage to its Building and Network,” December 28, 2020. https://www.datacenterfrontier.com/featured/article/11428484/nashville-explosion-at038t-repairs-damage-to-its-building-and-network

    [24] Tennessee Lookout, “Tennessee board details impact of Christmas bombing on 911 systems,” May 19, 2021. https://tennesseelookout.com/2021/05/19/tennessee-board-details-impact-of-christmas-bombing-on-911-systems/

    [25] Urgent Communications, “Tennessee board reports Nashville bombing impact on 911, future plans,” May 21, 2021. https://urgentcomm.com/911/tennessee-board-reports-nashville-bombing-impact-on-911-future-plans

    [26] APCO International, Public Safety Communications, “Nashville Christmas Bombing: Lessons Learned,” August 15, 2021. https://www.apcointl.org/psc/nashville-christmas-bombing-lessons-learned/

    [27] FOX 9, report on 911 service for Sprint customers in the Twin Cities after the Omaha fire, December 2016. https://www.fox9.com/news/911-service-restored-for-sprint-customers-in-metro-area

    [28] 3 News Now (KMTV), report on the Sprint outage and 911 in Omaha, December 2016. https://www.3newsnow.com/news/local-news/sprint-outage-impacts-911

    [29] Star Tribune, “CenturyLink weaknesses faulted in Minnesota 911 outage,” June 8, 2019. https://www.startribune.com/centurylink-weaknesses-faulted-in-minnesota-911-outage/510984822

    [30] Government Technology, “Widespread Minnesota 911 Malfunction Blamed on Bad Card,” January 13, 2021. https://www.govtech.com/public-safety/Widespread-Minnesota-911-Malfunction-Blamed-on-Bad-Card.html

    [31] Associated Press via Standard-Examiner, “Light pole installation hits fiber line, knocks out 911 service in several states,” April 18, 2024. https://www.standard.net/top-headlines/2024/apr/18/light-pole-installation-hits-fiber-line-knocks-out-911-service-in-several-states/

    [32] Cowboy State Daily, “Broken Fiber Line In Park County Exposes Fragility In Wyoming’s 911 System,” January 7, 2026. https://cowboystatedaily.com/2026/01/07/broken-fiber-line-in-park-county-exposes-fragility-in-wyomings-911-system

    [33] KTVQ, “911 service restored in northern Wyoming after fiber line vandalism,” January 9, 2026. https://www.ktvq.com/neighborhood-news/northern-wyoming/911-service-restored-in-northern-wyoming-after-fiber-line-vandalism

    [34] Economics and Technology, Inc., for the California Public Utilities Commission, Examination of the Local Telecommunications Networks and Related Policies and Practices of AT&T California and Frontier California, chapter 10, “Assessment of Safety, Redundancy and Resiliency of Network(s): Frontier,” 2019 (public version per D.20-12-021). https://www.cpuc.ca.gov/-/media/cpuc-website/divisions/communications-division/documents/network-exam-documents/network-exam-ch-10-redundancy-resiliency-frontrpublic-per-d2012021redacted.pdf

    [35] National 911 Program, California Office of Emergency Services NG911 Lessons Learned Report, November 7, 2022. https://www.911.gov/assets/National-911-Program_CalOES-Lessons-Learned-Report_7NOV2022_Final.pdf

    [36] Neb. Rev. Stat. § 86-1068. https://nebraskalegislature.gov/laws/statutes.php?statute=86-1068

    [37] Nebraska next generation 911 service contractor, State of NE 911 diversity audit (2Q24 snapshot), filed with the Legislature November 14, 2024. https://www.nebraskalegislature.gov/FloorDocs/108/PDF/Agencies/Next-Generation_911_Service_Contractor/855_20241114-094126.pdf

    [38] Nebraska Public Service Commission, 2022 Annual Report on Telecommunications. https://psc.nebraska.gov/sites/psc.nebraska.gov/files/doc/FINAL2022AnnualReportonTelecommunicationsFINAL2022-29-08.pdf

    [39] Nebraska Public Service Commission, Application No. C-4347, order on Nebraska Public Power District’s dark fiber lease to Pinpoint Communications, entered May 17, 2011. https://www.nebraska.gov/psc/orders/telecom/C-4347.pdf

    [40] Grant Schulte, “Towers and cable: These specialists help keep OPPD connected,” The Wire (Omaha Public Power District), June 1, 2026. https://oppdthewire.com/towers-and-cable-these-specialists-help-keep-oppd-connected/

    [41] Nebraska Department of Transportation, RFP 126331 Z6, Fiber Optic Cable Repair Services, Attachment B, fiber inventory. https://dot.nebraska.gov/media/exsbxlne/126331-z6-attachment-b-fiber-inventory.pdf

    [42] Hamilton Telecommunications, About. https://hamiltontel.com/about/

    [43] NTIA, Broadband Technology Opportunities Program fact sheet, NebraskaLink, LLC. https://www2.ntia.doc.gov/sites/default/files/grantees/fact_sheet_-_nebraskalink.pdf

    [44] Randy Bias, “The History of Pets vs Cattle and How to Use the Analogy Properly,” Cloudscaling. https://cloudscaling.com/blog/cloud-computing/the-history-of-pets-vs-cattle/

    [45] Montana Public Radio, “Flathead County 911 Suffers Thursday Morning Outage,” November 9, 2017. https://www.mtpr.org/montana-news/2017-11-09/flathead-county-911-suffers-thursday-morning-outage

    [46] Nebraska State Purchasing Bureau, RFP 202691101 solicitation page (bidders and documents). https://das.nebraska.gov/materiel/purchasing/RFP%20202691101/RFP%20202691101.html

    [47] Nebraska State Purchasing Bureau, notice of intent to award, RFP 202691101, July 30, 2026. https://das.nebraska.gov/materiel/purchasing/RFP%20202691101/2026.7.30%20911%20RFP%20Intent%20to%20Award.docx

    [48] Lumen, response to RFP 202691101, Technical File 1 of 4, section 3.1. https://das.nebraska.gov/materiel/purchasing/RFP%20202691101/202691101%20Lumen%20Technical%20File%201%20of%204.pdf

    [49] FCC, “FCC Modernizes Next Generation 911 Reliability and Interoperability,” news release, June 25, 2026. https://docs.fcc.gov/public/attachments/DOC-422582A1.pdf

    [50] 47 CFR § 4.9, Outage reporting requirements, threshold criteria (Legal Information Institute). https://www.law.cornell.edu/cfr/text/47/4.9

    [51] Indiana Statewide 911 Board, IN911 Network Overview. https://in911.net/network-overview.html

    [52] Starlink, business service plans, prices shown November 17, 2026. https://starlink.com/business/service-plans

    [53] Tanner Lee, HighSpeedInternet.com, “Our Starlink Business internet review,” August 7, 2026. https://www.highspeedinternet.com/providers/starlink/business

    [54] Rich Miller, Data Center Knowledge, “Cable Cut Cited in Silicon Valley Outage,” April 9, 2009. https://datacenterknowledge.com/cables/cable-cut-cited-in-silicon-valley-outage

    [55] ARRL, “When Vandals Strike Infrastructure, Hams Provide Communications Support,” April 15, 2009. https://www.arrl.org/news/when-vandals-strike-infrastructure-hams-provide-communications-support

    [56] InfoWorld, “Sandy slams mobile, wired and cable networks as far west as Michigan,” October 30, 2012. https://www.infoworld.com/article/2280967/sandy-slams-mobile-wired-and-cable-networks-as-far-west-as-michigan-2.html

    [57] California Public Utilities Commission, Safety and Enforcement Division, Incident Investigation Report, Incident No. E20130416-01, PG&E Metcalf Transmission Substation, June 24, 2013 (in PG&E’s January 30, 2015 filing, A.13-12-012). https://files.cpuc.ca.gov/PG&E20150130ResponseToA1312012Ruling/2014/02/SB_GT&S_0641100.pdf

    [58] Teresa Hansen, Renewable Energy World, “Wellinghoff Nailed it: PG&E Substation Attack is Terrorism,” May 14, 2014. https://www.renewableenergyworld.com/?p=8079817165

    [59] CBS San Francisco and Bay City News Service, “Vandalism At San Jose PG&E Substation Called ‘Sabotage’,” April 16, 2013. https://www.cbsnews.com/sanfrancisco/news/gunshots-cause-oil-spill-at-san-jose-pge-substation

    [60] FCC Public Safety and Homeland Security Bureau, March 8th, 2017 AT&T VoLTE 911 Outage Report and Recommendations, PS Docket No. 17-68, May 2017. https://docs.fcc.gov/public/attachments/DOC-344941A1.pdf

    [61] Telecommunications Law Professionals, “FCC Settles 911 Reliability Investigations with 4 Carriers,” December 20, 2021. https://www.tlp.law/2021/12/20/fcc-settles-911-reliability-investigations-with-4-carriers/

    [62] Nebraska Public Service Commission, “Service Restored to 911 Centers Impacted by Windstream Outage,” September 3, 2023. https://psc.nebraska.gov/node/1496

    [63] Nebraska Public Service Commission, “PSC Opens Investigations into 911 Outages,” 2023. https://psc.nebraska.gov/for-consumers/psc-opens-investigations-911-outages

    [64] Noah Crenshaw, Daily Journal (Franklin, Indiana), “Indiana State Treasurer: 911 outage was caused by server disruptions,” December 18, 2023. https://dailyjournal.net/2023/12/18/indiana-state-treasurer-911-outage-was-caused-by-server-disruptions/

    [65] FCC Public Safety and Homeland Security Bureau, February 22, 2024 AT&T Mobility Network Outage Report and Findings, July 22, 2024. https://docs.fcc.gov/public/attachments/DOC-404150A1.pdf

    [66] Associated Press via KIOS, “Telecom Outage Disrupts 911 Service Across IA,” May 31, 2024. https://www.kios.org/news/2024-05-31/telecom-outage-disrupts-911-service-across-ia

    [67] Iowa Department of Homeland Security and Emergency Management, “Interruption to 911 landline services reported throughout Iowa,” May 30, 2024. https://homelandsecurity.iowa.gov/hsem-news/2024-05-30/interruption-911-landline-services-reported-throughout-iowa

    [68] Massachusetts Executive Office of Public Safety and Security, “Massachusetts State 911 Department Determined Firewall Caused Statewide 911 Disruption,” June 19, 2024. https://www.mass.gov/news/massachusetts-state-911-department-determined-firewall-caused-statewide-911-disruption

    [69] 3 News Now (KMTV), “Lumen says a ‘fiber cut’ led to 911 disruptions in Southeast Nebraska counties,” May 29, 2025. https://www.3newsnow.com/news/local-news/lumen-says-a-fiber-cut-led-to-911-disruptions-in-southeast-nebraska-counties

    [70] Associated Press via WTOP, “AT&T attributes mass 911 outages in 3 states to fiber cuts made by ‘third parties’,” September 26, 2025. https://web.archive.org/web/20251008091618/https://wtop.com/national/2025/09/att-attributes-mass-911-outages-in-3-states-to-fiber-cuts-made-by-third-parties/

    [71] FCC Public Safety and Homeland Security Bureau, “Public Safety and Homeland Security Bureau Seeks Information on Effects of January 2026 Verizon Outage on Wireless Service Consumers, Public Safety Entities, and Government Entities,” Public Notice DA 26-94, PS Docket No. 26-21, January 28, 2026. https://docs.fcc.gov/public/attachments/DA-26-94A1.pdf

    [72] Jake Neenan, Broadband Breakfast, “Verizon Says Major Outage Caused by ‘Software Issue’,” January 20, 2026. https://broadbandbreakfast.com/verizon-says-major-outage-caused-by-software-issue/

    [73] Stephanie Weaver, FOX 5 DC, “Verizon outage: Users report stuck in SOS mode,” January 14, 2026. https://www.fox5dc.com/news/verizon-outage-users-report-stuck-sos-mode

    [74] Stateline, “FCC adopts new rules as states transition to Next Generation 911,” June 26, 2026. https://stateline.org/2026/06/26/fcc-adopts-new-rules-as-states-transition-to-next-generation-911/

    [75] KIRO 7, “911 service restored after Washington experiences statewide outage,” June 16, 2026. https://www.kiro7.com/news/local/911-service-restored-after-washington-experiences-statewide-outage/ZAWGQMPU2JC4ZEL5R7KMORPDS4/

    [76] MPR News, “Outage at Minneapolis data center behind last week’s telecommunications disruptions,” August 10, 2026. https://www.mprnews.org/story/2026/08/10/how-an-outage-at-unassuming-building-in-minneapolis-sent-waves-through-the-midwest

    [77] Regina Medina, MPR News, “Flights resume at Minneapolis-St. Paul International Airport after ‘telecommunication issues’,” August 6, 2026. https://www.mprnews.org/story/2026/08/06/flights-halted-at-minneapolisst-paul-international-airport

  • The FAA Leases a Fragile Network, and the November 13 Ground Stop Was the Latest Proof

    The FAA Leases a Fragile Network, and the November 13 Ground Stop Was the Latest Proof

    Mike Hammett has spent more than 20 years building and operating ISP, voice, internet exchange, and colocation networks.

    This article is part of a series on government network outages. The other pieces will be published as they are completed.

    TL;DR

    • The November 13, 2026, ground stop halted U.S. departures for more than four hours and disrupted communications at up to 17 of the FAA’s 22 en route centers [1][2]. The FAA’s initial investigation points to two remote sites, including one where L3Harris, its network contractor, was deploying a new router [2].
    • By this article’s count, it was at least the thirty-fourth FAA telecom or network failure since 1988 to take service down. In each of the 30 with a published cause, counting November 13’s preliminary one, one failure alone took service down, or two failures shared a cause, or a second failure came before the first was repaired.
    • The FAA leases its network, and the contract sets the design. In 2008, 16,400 of 21,871 planned services, including en route air-to-ground communications, were in a single-path class [3]. Replacing copper with fiber does not fix that.
    • The FAA can write diversity audits, staged changes, out-of-band management, and a third path that is not wireline into its new AT&T contract, and it should require the same of L3Harris, which runs today’s network, until the AT&T network replaces it.

    November 13, 2026

    Shortly after 3:30 p.m. Eastern on November 13, 2026, the FAA saw instability in radar and communications at its centers and terminal radar approach controls (TRACONs) [1]. At 3:54 p.m., it stopped departures nationwide [1]. The FAA later said instability began at 17 of its 22 Air Route Traffic Control Centers and spread into several TRACONs [2], and FAA Administrator Bryan Bedford said “no facilities went dark” [4]. The stop lifted at 8:15 p.m. [1]. FlightAware counted more than 9,000 U.S. flights canceled that day and about 6,700 the next [4], though the FAA says not all of the day’s cancellations were directly caused by the ground stop [2]. That day, Transportation Secretary Sean Duffy said the routers “got overloaded with data” and blamed “analog telecom” [1].

    At a briefing on November 14, Duffy said a preliminary investigation found that an L3Harris contractor was switching old copper wire to fiber at a remote FAA location while replacing a router [5]. The router “started flooding other routers in our air traffic control system,” he said, with “over 1,000 connection attempts every four to five seconds” [4][6]. CBS reported that the router flooded the system for 45 minutes and that the FAA could not determine the cause of the instability [5]. “Once we’re able to isolate what was being done at this remote location, all of our systems came back online,” Duffy said [4]. Nick Daniels, who heads the controllers’ union, said, “We know we didn’t have communications with airplanes, or that we were switching airplanes to different frequencies because their communications were being disrupted” [4].

    Duffy put the open question as whether it was “user error or install error from L3Harris” or a faulty router, and said, “We haven’t determined what the fault was” [6]. He said that “L3Harris actually owns that network” [7] and that he did not believe the outage was a cyberattack [4][6]. Asked why the work was done at 3 p.m. on a Friday, he said, “It shouldn’t have that kind of impact,” and that there would be “a reevaluation of when we do this work” [4][6]. In May, when L3Harris announced it had passed the halfway mark of its rebuild of the FAA network, it said the work was being done with “zero operational disruptions” [8].

    On November 15, The Wall Street Journal reported that the router was a new Juniper Networks device L3Harris had installed at an FAA radio site in El Dorado, Arkansas, and that data from it, and possibly from a router at a radar site in Amarillo, Texas, overwhelmed the system [9]. The same day, Duffy said the Amarillo router, replaced about three weeks earlier, had also flooded the network, and that both routers were at an L3Harris lab in Florida for testing with engineers from the FAA, Juniper, and Cisco [10][11]. He put the number of routers of that type at 69 [10], and the Journal reported about 60 Juniper routers in the FAA’s network [9]. Officials identified a router feature that appeared to cause the overload, which a government official described as a broadcasting feature that sent data “more broadly than necessary,” and the FAA has disabled it on all its Juniper routers [9]. A spokesman for Hewlett Packard Enterprise, which owns Juniper, said the devices did not fail and appeared to have been working as intended [9]. After CenturyLink’s 2018 outage, which spread over a management channel that was enabled by default and unused, the FCC concluded that system features not in use “should be turned off or disabled” [12].

    Duffy is right that the work shouldn’t have that kind of impact, at any hour. Night work would have put fewer aircraft in the air when it failed, but it would not have kept the trouble at the sites where it started. More than 1,000 connection attempts every four to five seconds is roughly 200 to 250 a second. The reason for outage should say what kind of traffic that was and why it was enough to destabilize the system, whether control-plane policing or storm control was configured on these routers, whether the feature the FAA switched off was on by default, and whether L3Harris could reach the affected routers over a path that did not run through them.

    Two months before the outage, GAO concluded that “the FAA is buying delivery speed at the risk of cost uncertainty and potential disruptions,” and warned that installations scheduled at the same sites without addressing conflicts risked “excessive disruptions to air traffic control operations” [13]. With 5,170 lines to cut over in all, every cutover is a change on a live network, and the program has to make each one unable to reach past its own site.

    The FAA has published a preliminary account but not a final reason for outage, and it says it and L3Harris “continue to investigate the root cause” [2]. Duffy said the findings would be shared when the review is done, and Bedford said he expected “a very transparent hearing” [4]. Representatives Rick Larsen and André Carson, the top Democrats on the House Transportation and Infrastructure Committee and its aviation subcommittee, said they expect to hear from Bedford at a hearing by early December [14]. On a properly designed network, a change reaches one half of a redundant pair, proves itself, and only then reaches the other half. A change that can reach every center in one step is a single point of failure, however old the routers are.

    The FAA Leases Its Network

    The FAA Telecommunications Infrastructure (FTI) program, awarded in 2002, put the agency’s communications on a contracted network [15]. The DOT Inspector General described FTI in 2015 as owned and maintained by Harris Corporation and leased to the FAA [16].

    The successor program, FAA Enterprise Network Services (FENS), went to Verizon Business Network Services on March 27, 2023 [17], with a 15-year value of $2.4 billion [15][17]. On August 11, 2026, the FAA used waiver authority to make a direct award to AT&T Enterprise Services [18], with an initial $74.3 million task order while the long-term contract is negotiated [19][20]. Verizon moved to a supporting role [19]. FTI still carries the FAA’s traffic: the FAA’s account of November 13 says its legacy equipment became unstable when data began flooding “the FAA FTI network” [2].

    That makes these outages failures of carrier-grade network engineering, delivered by carriers and defense contractors under federal contracts. The FAA’s part is what it specifies, what it pays for, and what it verifies. The public record does not show whether the FAA now verifies physical path diversity on its leased circuits.

    1988 to 1991: The Template

    On May 8, 1988, fire broke out at Illinois Bell’s Hinsdale central office, west of Chicago [21]. The circuits between Chicago Center in Aurora and O’Hare’s tower ran through that one building, and the fire crippled the center’s communications with O’Hare and Midway [22][23]. Controllers at both airports widened the spacing between flights to 20 miles, and O’Hare, which could handle 150 aircraft an hour, took only 50 arrivals an hour [22].

    On May 4, 1991, four of the FAA’s major air traffic control centers shut down for 5 hours and 22 minutes, and the cause was logged as “Fiber cable cut by farmer burying dead cow. Lost 27 circuits” [24]. Later that year, in a report to a House subcommittee, the FAA counted 114 major telecommunications outages that led to flight delays between August 1990 and August 1991 [24].

    On September 17, 1991, AT&T moved its Thomas Street switching center in lower Manhattan from Con Edison power to its own diesel generators under a load-shedding agreement. A voltage surge during the transfer disabled the equipment powering the transmission facilities and the alarm system with it, according to the FCC’s report [25]. A burned-out indicator lamp, a muted alarm bell, and a power crew sent to an off-site training class meant nobody noticed until the batteries were nearly drained [26]. GAO reported in 1993 that the outage caused a significant loss of the FAA’s radar sites, radio channels, and computer links, along with 1,174 canceled or delayed flights [27].

    2009: One Router Error, 129 Facilities

    On November 19, 2009, an FTI routing failure took down flight plan processing nationwide from about 5:15 a.m. to 9 a.m. Eastern. The failure also took down the second national node in Hampton, Georgia [28]. The National Airspace Data Interchange Network (NADIN) ran there and in Salt Lake City, so the geographic redundancy between Utah and Georgia did not help [28].

    The DOT Inspector General’s review of the 2009 outage traced the cause to Harris. During FTI’s transition from the old network to the new fiber network, Harris had installed temporary “bridge” routers at 26 core sites. On October 21, 2009, an engineer misconfigured the one at Los Angeles Center, and the error sat dormant for four weeks. On November 19, an engineer put a replacement router in service at Los Angeles and copied over a route map that lacked a needed “no” statement [29]. The error blocked about 75 percent of routes across the FTI fiber network for five hours. It reached 129 facilities instead of the 21 it should have affected, and it delayed 820 flights [29].

    Restoration took five hours partly because Harris’s automated alert tool was misconfigured, and its filter “caused the alarms to be filtered for all routers on the FTI network” [29]. When the fix came, it waited on a technician to reach Salt Lake City and pull a router card [30].

    The Inspector General’s letter also covered what the outage exposed. Harris had confirmed “either the complete lack or inadequate proof of diversity” between primary and alternate paths at several critical facilities, including the FAA’s William J. Hughes Technical Center and the Baltimore tower, and some FTI sites had lost their diversity after AT&T or Sprint upgraded their networks [29]. FAA had three representatives on site at Harris but had “limited ability to oversee FTI,” could not directly monitor the FTI network, and took a reactive approach to network vulnerabilities [29]. As of December 2009, 8.19 percent of FTI services did not meet their availability requirements, and the FAA’s Automatic Dependent Surveillance-Broadcast (ADS-B) program had chosen not to use FTI because of reliability and security concerns [29].

    One latent error, introduced during a network transition, hidden by the alarms meant to catch it, and copied into new equipment, reached six times as many facilities as it should have.

    2014: Two Fiber Paths Into One Room

    On September 26, 2014, an FAA contract employee deliberately started a fire that destroyed FTI equipment at Chicago Center in Aurora, Illinois [16][31]. The center could not control traffic for more than two weeks [16].

    Outside the building, FTI had two independent primary and backup fiber pathways. Inside, all FTI equipment racks, primary and backup, were located close to one another [16]. Water from the fire suppression system destroyed about half of them [16]. Chicago TRACON lost its long-range radar because that feed was routed through Chicago Center [16]. Rerouting communications to other facilities took days rather than hours, and the backup facility could not be used because its equipment was incompatible [16]. Diverse fiber that lands in one room is a single point of failure.

    Newark and Philadelphia: The Same Facility Three Times

    The FAA moved Newark approach control from the New York TRACON to the Philadelphia TRACON in 2024 [32]. Radar data for Newark still came from an FAA facility in Westbury, on Long Island, over a remote line to Philadelphia [33]. On April 28, 2025, controllers working Newark lost radar and radio for about 90 seconds because of a burnt copper wire, and it happened again on May 9, when primary and redundant communication infrastructure both failed [34]. Duffy said that on April 28, “the primary communication line went down, the backup line didn’t fire” [35]. GAO later reported the FAA’s own attribution, which named old copper wiring and moving the airspace “without adequately testing the reliability of the commercial telecommunications lines used for critical data” [13].

    The FAA said later that on September 20, 2026, after a series of error messages on a telecommunications circuit, it had moved the Philadelphia TRACON onto its backup fiber at the request of its contractor, L3Harris, so the primary could be troubleshot [36]. The next morning, a contractor on NJ Transit’s Delco Lead project severed that backup near New Brunswick at about 9:45 a.m. [36][37]. Bedford called the primary one of the “old L3 Verizon circuits” [38], and Verizon said its fiber running beside the rail line had been cut [39]. More than 600 flights into and out of Newark alone were canceled [40]. Two days later, Bedford said replacing the remaining copper with fiber would not be enough to prevent future outages [36].

    L3Harris’s FTI page says the network uses multiple local exchange carriers and that the approach “ensures redundancy and reliability across the system” [41]. That describes the network as a whole. At Philadelphia, both paths were Verizon facilities [38][39]. Taking one half of a pair out of service makes the other half a single point of failure until the first comes back, and that window needs its own protection. A same-carrier backup left as the only path, with nothing standing by behind it, breaks the rules of a properly designed network.

    Dallas 2025: Primary and Secondary, Cut Together

    On September 19, 2025, the FAA’s Dallas TRACON lost radar, frequencies, and automation, and both Dallas airports went into ground stops for hours [13][42]. The FAA blamed “multiple failures of the TDM data telecommunications service provided by Frontier” [42]. American Airlines said the FAA told it two cut fiber cables affected the primary and secondary data paths [43]. Frontier said a third-party contractor in Argyle, Texas, cut its fiber lines, and L3Harris said a fault in the local carrier’s switching gear and a fiber cut both played a part [44]. The FAA itself said an oversight by L3Harris failed to make sure the redundancies in the system functioned properly [43]. More than 1,800 flights were delayed or canceled [13].

    By the companies’ own accounts, one afternoon of damage inside one local carrier’s network reached both the primary and secondary paths for the TRACON serving both Dallas airports [43][44]. Whether the two paths also shared a trench, a carrier office, or a TDM switch is what a diversity audit exists to find.

    Old Copper Is the Wrong Diagnosis

    The FAA attributed Newark to old copper wiring [13], and on November 13 Duffy blamed “analog telecom” [1]. Asked whether its upgrades and modernization efforts caused the ground stop, the FAA said no, that it was “due to the legacy TDM system causing instabilities” [2]. The copper is from the 1960s, the technology is no longer supported by the carriers, and modernization needs money [13]. Congress approved $12.5 billion in 2025 [45], and after the November 13 ground stop, Duffy raised the administration’s request to $50 billion [46]. As of May 2026, the FAA had replaced 2,560 of 5,170 copper lines with fiber [13].

    Age explains why components fail more often. It does not explain why a cut on the only working path stops five airports. The Philadelphia backup that was cut in September was fiber, and so were the Dallas paths. On November 13, the change that went wrong was a new router installed during the copper replacement, and data from routers at two sites overwhelmed the system [5][9][10]. Duffy said the telecom provider had forced the move from copper to fiber, and that the cutover work had been done thousands of times and “this has never happened” [47]. Transition work took most of the network down in 2009, when a replacement router put in service during FTI’s move to a new fiber network blocked about 75 percent of the network’s routes [29]. Duffy said on November 15 that disruptions could continue until the last copper lines are replaced at the end of 2028 [9]. A new fiber circuit from the same carrier, in the same right-of-way, landing in the same room, inherits the single points of failure of the copper it replaced. Funding without a diversity requirement buys newer single points of failure.

    The Contract Is the Design

    When an agency leases its network, the contract decides how diverse it has to be. The DOT Inspector General’s 2008 audit of FTI laid this out. FTI services were sorted into reliability classes. Classes RMA-1 through RMA-3 were built on dual paths. RMA-4 and RMA-5 were single-path, and RMA-4, which included en route air-to-ground communications, made up 16,400 of 21,871 planned services [3]. RMA-4 services had to be restored within 3 hours. In March 2008, the contractor averaged 5 hours and 47 minutes [3].

    The same audit found that the Kennedy tower, the New York TRACON, and New York Center had all of their FTI services routed through one switching center run by one telephone company, with no alternate [3]. It found the September 25, 2007, Memphis Center outage, which caused 566 delays over 3 hours, would have been prevented if the alternate path had been routed differently [3]. The audit also counted 21 unscheduled FTI outages with flight delays between December 2006 and August 2007 [3]. FAA told the auditors that FTI was already performing regularly scheduled diversity audits [3]. The contract’s remedy for a missed specification was credits paid back to the FAA [3]. A credit refunds part of the bill. It does not move a circuit out of a shared conduit.

    The FAA made its August 2026 award to AT&T without a new competition. Its justification called AT&T the “only Tier-1 provider currently willing and able” to meet the program’s timeline and said a competitive procurement “would require a minimum of 22 months,” too long for the December 2028 deadline [18][20]. It defined a Tier-1 provider as one that “owns core network infrastructure” and said direct engagement with one provides “superior network resiliency” [18]. A buyer that has told one carrier it is the only option has less bargaining power to demand diversity audits, tagging, and penalties. In the industry, Tier-1 describes a network that reaches the whole internet without paying anyone for transit. Neither definition says anything about whether two paths into a center share a trench.

    The chain of middlemen makes all of it harder to verify. Under FTI, the FAA contracts with L3Harris; L3Harris buys circuits from Verizon, Frontier, and other local carriers; and those carriers’ outside plant is dug up by other parties’ contractors [39][41][44]. Under FENS, AT&T is prime, Verizon still runs core network functions, and Peraton is the integrator for the wider overhaul [19].

    Chaining carriers is a legitimate way to build a diverse route, and it holds up only when someone maps the whole chain to the trench and checks it again after every change. The party that owes the FAA diversity is rarely the party that holds the conduit maps, so every layer between the agency and the trench is another place where a regroomed circuit goes unnoticed. The Defense Department’s rule for its own sites requires the two circuits that connect a site to its core network to stay separate along their entire length, end at two geographically diverse core nodes, and use different “cables, demarks, and digital cross connects” [48].

    What a Properly Designed Network Does

    Components fail. Fiber gets cut, rectifiers trip, line cards die, and people push bad configurations. A properly designed network accepts that and never lets one failure become a service outage. Two independent failures can still overlap, but two failures that share a cause are a design defect. A “diverse” circuit in the same cable, the same conduit, the same carrier office, or the same equipment room was never diverse. Failures also line up in time, when a backup fails unnoticed or a failed half of a pair stays out of service, and a redundant design runs single-threaded until someone fixes it.

    In a properly designed network, primary and backup paths run in separate conduits and through separate building entrances, from different carriers or with documented proof that they share no facility, into equipment in different rooms. The backup carries live traffic or is tested continuously, and a failed half of a pair is repaired as an outage in progress. Every critical device can be reached out of band, over a path that shares nothing with the network it manages. Changes go to one half of a pair at a time, and the reach of any one failure stays small enough that a cut, a card, a power plant, or a change in one place takes down service only where it sits.

    The FAA has had its own diversity order since November 1995, when it issued Order 6000.36A, “Communications Diversity”; it was replaced in November 2021 by Order JO 6000.36B, “National Airspace System Service Survivability” [49]. The FAA’s own radio communications order requires engineering for “diversity, redundancy and no single point of failure” based on service criticality [50]. The failures in this record that have a published cause are what those orders are written to prevent.

    The Strongest Case for the FAA

    The best defense is scale. FTI connects about 5,000 FAA facilities, radars, voice systems, and flight data links, and controllers manage up to 45,000 flights a day [13]. Ground stops are the system failing safe, which is what it should do. Some FAA sites sit where only one carrier or one conduit exists, and true diversity there costs real money.

    The ground stops were the right call. Still, at least 29 of the 34 events in the table hit an en route center, a TRACON, a major airport, or the national flight-plan system, and the 2008 audit gives only a location for the other five. The FAA has 22 en route centers [2]. Those buildings and the large TRACONs carry the national system, and most of them sit in or near large metro areas, where a second carrier or a second route is easiest to buy. The 2015 Inspector General report already said the limits seen in Chicago would be similar at many of the centers [16].

    What Would Change the Record

    The first step is for the FAA to publish a reason for outage for every failure that stops traffic, with an initial report within 72 hours and a final report within 30 days, the same clock its carriers already follow with the FCC [51]. Its carriers already have to file outage reports with the FCC when a primary and backup both fail at a center or airport for at least 30 minutes [51][52], and the FCC’s rules presume those reports are confidential [53]. The public record for most of these events is a press conference and a few sentences. Officials say November 13 began inside L3Harris’s own routers. If so, no carrier may have had an outage to report, and the public record will depend on what the FAA and L3Harris disclose, at the hearing and after.

    On August 6, 2026, the FAA paused flights into and out of Minneapolis Center’s airspace over what it called telecommunications issues, and the airport said the ground stop at Minneapolis-St. Paul ran from about 2 p.m. to 4 p.m. [54]. The same day, a power failure at the 511 Building, a downtown Minneapolis carrier hotel and data center, reached the building’s backup power [55]. MPR News reported it was unclear whether the same failure caused the flight pause [55]. The FAA has not said what failed, so the public cannot tell. If the link holds, it is 1991 again, with a federal safety service riding through a single building’s power plant.

    The DOT Inspector General suggested in 2010 that the FAA require independent periodic reviews of its network architecture [29], and publishing the findings would let the public see whether the fixes held. The reviewer should be a network engineer who works for the FAA or for a firm with no stake in the contract, and who writes the physical requirements, reads the route maps, and witnesses the failover tests.

    The second is to hold FAA circuits to at least the annual audit standard the FCC set for 911 carriers in 2013, which its June 2026 order still accepts from legacy providers even as it ends the annual certification [56], and to the diversity audits the FCC has urged on every provider since at least 2020 [57]. That means an annual physical diversity audit of every critical circuit feeding a center or major TRACON, tagging so carriers cannot regroom those circuits onto a shared route without notice, re-audit after any rearrangement, and certification by an accountable official. The audit has to cover the power behind each path, including where the generators sit and who sees the battery alarms, because Thomas Street and AT&T’s Nashville office both failed on power [25][58].

    The third is to write the rest of the design into the contract. The August award to AT&T was undefinitized, with the long-term contract still to be negotiated [18][19]. Terms written into that award reach the FAA’s traffic only as it moves to FENS, so the same requirements belong in L3Harris’s FTI work until then. Carrier diversity, separate building entrances, separate equipment rooms, continuous monitoring of backup paths, and alarms that reach the FAA directly can all be requirements. So can out-of-band management that does not ride the network it manages, limits that keep one site’s routing or management failure at that site, and changes staged one half at a time, with remedies that go beyond service credits. New equipment can also be required to pass tests on a lab network that mirrors production, with unused features turned off, before it reaches the live network. Work that takes one half of a pair out of service should start only after the other half is verified, and that half should be protected until the work is done. Repair times belong on the redundant half as well as the primary, since the Telecommunications Service Priority program sets only the order of repairs [59]. A cap on how many facilities any single element can affect would turn the 129-instead-of-21 problem from 2009 into a contract term [29]. The cap only means something if the contractor has to prove it, by failing elements on a schedule and after major changes, showing the FAA what actually went down, and fixing any domain that turns out larger than the drawing.

    For the centers and the largest TRACONs, two wireline paths are not enough. A third path that is not wireline at all, such as licensed microwave or satellite, cannot share a trench, a conduit, or a central office with the other two. It only has to carry the critical minimum, the controllers’ voice to their radio sites and the radar feeds, but it has to be independent in every other way, with its own building entrance, power, carrier, and management. On November 13, a third path that rode FTI’s routers would have shared their failure. In 2018, the FTI program proposed commercial microwave to give the long-range radar at Elwood, Illinois, “telecommunications diversity,” replacing the FAA’s own 2 GHz microwave link [60]. In July 2025, Bedford told lawmakers the FAA would not rely on Starlink “or any satellite service” as the sole technology for safety-critical air traffic services, and that FTI was testing a few dozen Starlink terminals [61]. A third path earns its place only if it fails for reasons the other two do not share. One satellite network can serve every center on those terms, because an outage of that network leaves the two wireline paths up.

    The fourth is to finish what the 2015 report asked for, so that any center’s airspace and data can move to a neighboring center in minutes [16]. After the Chicago fire, the center’s automation could not hand its airspace to a neighbor automatically, and the workaround took days [16].

    Quick Compare: The FAA Record

    The table counts every FAA telecom or network failure since 1988 that this article found documented and that took a service down, even briefly. Rows without a published cause stay in the count, and the pattern claim in the TL;DR covers only the rows that have one. The last column is this article’s reading of each cited cause. Where a row’s source gives only the trigger, the reading is that one trigger was enough to take the service down.

    DateServiceReported causeWhat was single
    May 8, 1988Chicago Center lines to O’Hare and MidwayHinsdale central office fire [21][22]One central office
    October 15, 1990O’Hare tower, delays up to two hoursTree-planting contractor cut a line [62]One cable
    About November 1990Kansas City air traffic center, 4 hours 16 minutesBeaver chewed a fiber cable [24]One cable
    January 4, 1991New York ATCAT&T crew cut fiber in Newark [63]One cable route
    May 4, 1991Four en route centers, 5 hours 22 minutesFarmer cut a fiber cable [24]One cable
    September 17, 1991New York ATC, 1,174 flightsPower transfer, alarms missed, batteries ran flat [25][26][27]One building’s power plant
    September 28, 2006Atlanta tower, 8 delaysDefective equipment [3]One equipment failure
    December 4, 2006San Juan Center, 20 hours, 60 delaysUnderwater cable cut [3]One cable
    December 11, 2006Orlando tower, 63 delaysImproper configuration [3]One configuration error
    January 9, 2007Salt Lake Center, 92 delaysMaintenance procedures not followed [3]One maintenance error
    March 30, 2007Los Angeles tower, 25 delaysFTI troubleshooting [3]Troubleshooting work
    May 23, 2007Chicago, 77 delaysMisconfigured FTI equipment [3]One configuration error
    July 27, 2007FTI en route communications over North CarolinaPrimary and backup lost; no cause stated [3]Unknown
    September 25, 2007Memphis Center, 566 delaysInadequate physical diversity [3]Alternate path routing
    November 9, 2007Jacksonville Center, 85 delaysInadequate physical diversity [3]Primary and alternate together
    December 4, 2007Merrimack, New Hampshire, 49 delaysCables dislodged from a patch panel [3]One patch panel
    March 24, 2008Leesburg, Virginia, 60 delaysNo cause identified [3]Unknown
    April 12, 2008San Diego, 5 delaysTelco failure [3]One carrier
    May 15, 2008Chamblee, Georgia, 41 delaysFTI circuit failure [3]One circuit
    August 2008Memphis Center communications with aircraftCar hit a pole and cut fiber [64]One cable
    November 19, 2009Flight plans, nationwide, 129 facilitiesDormant router misconfiguration, alarms filtered [29]One change, one alarm filter
    September 26, 2014Chicago Center, 2+ weeksArson in telecom room [16]One room for primary and backup
    September 20, 2017FAA San Juan Center, Puerto Rico radarsHurricane Maria destroyed a key generator; two FAA radars inoperable [65][66]One generator
    December 25, 2020Memphis Center frequenciesNashville bombing cut central office power and generators [58][67]One central office
    April 28, 2025Newark approach via Philadelphia TRACON, about 90 secondsBurnt copper wire; backup line did not take over [34][35]Backup did not take over
    May 9, 2025Newark approach via Philadelphia TRACON, about 90 secondsPrimary and redundant infrastructure both failed [13][32][34]Untested commercial lines
    May 12, 2025Denver Center sector frequenciesBoth transmitters for a sector failed; cause unpublished [68]Unknown
    June 20, 2025Jacksonville Center radar, brieflyFiber line cut; backup kicked in [69]One fiber line, until the backup took over
    September 19, 2025Dallas TRACONPrimary and secondary paths cut [42][43]One local carrier
    August 6, 2026Minneapolis Center airspaceTelecom issue, cause unpublished [54]Unknown
    September 21, 2026Philadelphia TRACON, 5 airportsPrimary out for troubleshooting, Verizon backup cut [36][39]One working path, one carrier
    October 12, 2026Raleigh-Durham arrivals, hours of restrictionsPlumbing contractor cut fiber [70]One cable cut
    October 13, 2026Houston Intercontinental communicationsConstruction contractor damaged fiber [71]One cable cut
    November 13, 2026Up to 17 of 22 en route centersNew routers at two sites flooded the network, one installed during copper-to-fiber work (preliminary) [5][9][10]One router feature, network-wide

    Cover photo: WhisperToMe / Wikimedia Commons, public domain

    Sources

    [1] ABC News, “Nationwide ground stop lifted after more than 4 hours as Trump administration blames old, overloaded tech,” November 13, 2026. https://abcnews.com/US/nationwide-ground-stop-ordered-airports-faa/story?id=138121848

    [2] FAA, “Nationwide Ground Stop: Questions and Answers,” November 15, 2026 (the address of the FAA’s November 13 statement now redirects here). https://www.faa.gov/newsroom/nationwide-ground-stop-questions-and-answers

    [3] DOT Office of Inspector General, FAA’s Progress and Challenges in Meeting FTI Transition Goals, Report AV-2008-089, September 30, 2008. https://govinfo.gov/content/pkg/GOVPUB-TD-PURL-gpo27442/pdf/GOVPUB-TD-PURL-gpo27442.pdf

    [4] Yahoo News, “FAA ground stop latest updates,” live coverage, November 14 and 15, 2026 (headline updated during coverage). https://www.yahoo.com/news/us/live/faa-ground-stop-latest-updates-duffy-blames-contractor-l3harris-thousands-more-flights-canceled-saturday-no-sign-of-cyberattack-161329174.html

    [5] CBS News, “Travelers stranded and thousands of flights canceled after FAA ground stop. Here’s what we know,” updated November 15, 2026. https://www.cbsnews.com/amp/news/faa-ground-stop-canceled-flights-travelers-what-we-know/

    [6] U.S. Department of Transportation, “ICYMI: U.S. Transportation Secretary Sean P. Duffy Provides Update on Restoration of Airspace Following Nationwide Ground Stop,” November 14, 2026. https://www.transportation.gov/briefing-room/icymi-us-transportation-secretary-sean-p-duffy-provides-update-restoration-airspace

    [7] Fox Business, account of Duffy’s November 14 briefing on L3Harris, November 14, 2026. https://www.foxbusiness.com/fox-news-air-space/who-l3harris-federal-contractor-sec-duffy-says-may-have-caused-faa-outage

    [8] L3Harris, “L3Harris Reaches Over 50% in FAA Telecommunications Modernization,” May 11, 2026. https://www.l3harris.com/newsroom/editorial/2026/05/l3harris-reaches-over-50-faa-telecommunications-modernization

    [9] Andrew Tangel and Drew FitzGerald, “The Single Router in Southern Arkansas That Ground Air Travel to a Halt,” The Wall Street Journal, November 15, 2026 (republished by To Vima). https://www.tovima.com/wsj/the-single-router-in-southern-arkansas-that-ground-air-travel-to-a-halt

    [10] Quinn Scanlan, ABC News, “Transportation Secretary Duffy says ATC routers’ ability to broadcast data turned off, being investigated,” November 15, 2026. https://abcnews.com/Politics/transportation-secretary-duffy-atc-routers-ability-broadcast-data/story?id=138155545

    [11] FOX 4 Dallas-Fort Worth, “FAA ground stop: Routers in Texas, Arkansas being tested after outage,” November 15, 2026. https://www.fox4news.com/news/faa-ground-stop-routers-texas-arkansas-being-tested-after-outage

    [12] FCC Public Safety and Homeland Security Bureau, December 27, 2018 CenturyLink Network Outage Report, August 19, 2019. https://docs.fcc.gov/public/attachments/DOC-359134A1.pdf

    [13] GAO, Air Traffic Control Systems: Ambitious New Modernization Effort Needs to Improve Cost and Schedule Planning, GAO-26-107992, September 14, 2026. https://files.gao.gov/reports/GAO-26-107992/index.html

    [14] CNN, “US flight ground stop latest,” live coverage, November 15, 2026 (headline updated during coverage). https://www.cnn.com/2026/11/15/politics/live-news/trump-administration-ground-stop

    [15] Washington Technology, “FAA selects Verizon for $2.4B telecommunications contract,” March 2023. https://www.washingtontechnology.com/companies/2023/03/faa-selects-verizon-24b-telecommunications-contract/384517/

    [16] DOT Office of Inspector General, FAA’s Contingency Plans and Security Protocols Were Insufficient at Chicago Air Traffic Control Facilities, Report AV-2015-112, September 29, 2015. https://www.durbin.senate.gov/imo/media/doc/FAA%20contingency%20plans%20and%20security%20protocols%20at%20Chicago%20ATC%20facilities.pdf

    [17] SAM.gov, FAA Enterprise Network Services (FENS) award notice 693KA8-23-D-00010. https://sam.gov/opp/7e58b8f050ef40f59761aa3b334528a5/view

    [18] SAM.gov, FAA Enterprise Network Services (FENS) award notice 693KA8-26-D-00016 to AT&T, contract award date August 11, 2026, published August 12, 2026. https://sam.gov/opp/89275702a17c4bc4a1906fd596fe8828/view

    [19] Washington Technology, “FAA brings in AT&T for network modernization program,” August 2026. https://www.washingtontechnology.com/contracts/2026/08/faa-brings-t-network-modernization-program/415412/

    [20] MeriTalk, “FAA Awards AT&T $74M to Kick Off Network Overhaul,” August 14, 2026. https://meritalk.com/articles/faa-awards-att-74m-to-kick-off-network-overhaul/

    [21] Office of the State Fire Marshal and Illinois Commerce Commission Staff, Hinsdale Central Office Fire, Final Report, Executive Summary, March 1989. https://archive.decromancer.ca/vtda.org/docs/telephony/Ameritech/The_Hinsdale_Challenge_1988_Executive_Summary.pdf

    [22] UPI, “Bell puzzles over cause of crippling blaze,” May 9, 1988. https://www.upi.com/Archives/1988/05/09/Bell-puzzles-over-cause-of-crippling-blaze/6390579153600/ Archived copy: https://archive.ph/9A19S

    [23] TELECOM Digest, May 1988 issues on the Illinois Bell Hinsdale fire (archive). https://bio.nic.funet.fi/pub/mirrors/lcs.mit.edu/telecom-archives/TELECOM_Digest_Online/1309.html

    [24] RISKS Digest, Volume 12, Issue 62, November 12, 1991 (quoting the Associated Press, November 10, 1991). https://catless.ncl.ac.uk/Risks/12/62

    [25] UPI, “FCC orders AT&T to study communications system in wake of outage,” November 8, 1991. https://www.upi.com/Archives/1991/11/08/FCC-orders-ATT-to-study-communications-system-in-wake-of-outage/7189689576400/ Archived copy: https://archive.ph/mSFHD

    [26] UPI, “AT&T discloses cause of power failure,” September 20, 1991. https://www.upi.com/Archives/1991/09/20/ATT-discloses-cause-of-power-failure/8379685339200/ Archived copy: https://archive.ph/20261116200152/https://www.upi.com/Archives/1991/09/20/ATT-discloses-cause-of-power-failure/8379685339200/

    [27] GAO, Telecommunications: Interruptions of Telephone Service, RCED-93-79FS, March 5, 1993. https://www.gao.gov/assets/rced-93-79fs.pdf

    [28] eWeek, “FAA Flight-Plan System Crashes Again, Delays Hundreds of U.S. Flights,” November 19, 2009. https://www.eweek.com/storage/faa-flight-plan-system-crashes-again-delays-hundreds-of-u.s.-flights/

    [29] DOT Office of Inspector General, letter to Chairmen Oberstar and Costello on the November 19, 2009, FAA Telecommunications Infrastructure outage, June 17, 2010. https://www.oig.dot.gov/sites/default/files/WEB%20FILE2_FTI%20%20Nov%2019%20outage-6-18-10_Final%20Signature%28a%29%20%282%29.pdf

    [30] eWeek, “The Story Behind the FAA Flight-Plan System Crash,” November 20, 2009. https://www.eweek.com/networking/the-story-behind-the-faa-flight-plan-system-crash

    [31] CBS News, “Report slams FAA response to fire at Chicago air traffic facility,” September 30, 2015. https://www.cbsnews.com/news/report-slams-faa-response-to-fire-at-chicago-air-traffic-facility

    [32] Flying, “Newark Airport Suffers Another 90-Second Radar Outage,” May 2025. https://www.flyingmag.com/newark-airport-suffers-another-90-second-radar-outage/

    [33] Pete Muntean, Rene Marsh, TuAnh Dam, and Amanda Musa, CNN, via KVIA, “Newark Airport systems outage was ‘most dangerous situation,’ air traffic controller says,” May 6, 2025. https://kvia.com/cnn-national/2025/05/06/newark-airport-systems-outage-was-most-dangerous-situation-air-traffic-controller-says/

    [34] DOT Office of Inspector General, Audit Announcement, FAA’s Relocation of Newark-Area Terminal Radar Approach Control (TRACON) Responsibilities, Project No. 25A3005A000, July 28, 2025. https://www.oig.dot.gov/sites/default/files/library-items/Audit%20Announcement%20-%20FAA%20N90%20Review%207.28.25.pdf

    [35] Andrea Margolis and Grady Trimble, Fox Business, “Newark air traffic controllers lost radar, radio communications with planes for over a minute, sparking chaos,” May 5, 2025. https://www.foxbusiness.com/fox-news-travel/newark-air-traffic-controllers-lost-radar-radio-communications-planes-minute-sparking-chaos

    [36] Reuters via WHBL, “US FAA says error messages in telecommunications circuit prompted shift to backup on Sunday,” September 23, 2026. https://whbl.com/2026/09/23/us-faa-says-error-messages-in-telecommunications-circuit-prompted-shift-to-backup-on-sunday/

    [37] Airways, FAA Northeast ground stops and NJ Transit statement, September 2026. https://airwaysmag.com/new-post/faa-northeast-ground-stops-severed-fiber-philadelphia-tracon

    [38] AirlineGeeks, “ATC Outage Cancels, Delays Hundreds of Flights on East Coast,” September 21, 2026. https://airlinegeeks.com/2026/09/21/atc-outage-cancels-delays-hundreds-of-flights-on-east-coast/

    [39] Newsweek, Philadelphia TRACON outage and Verizon statement, September 21, 2026. https://www.newsweek.com/hundreds-of-flights-impacted-at-major-us-airports-over-equipment-failure-12469404

    [40] Associated Press via WSMV, “Flights disrupted across US Northeast due to technical problems at key air traffic center,” September 21, 2026. https://www.wsmv.com/2026/09/21/flights-disrupted-across-us-northeast-due-technical-problems-key-air-traffic-center/?outputType=amp

    [41] L3Harris, “FAA Telecommunications Infrastructure (FTI),” service page. https://www.l3harris.com/all-capabilities/fti-network-managed-services

    [42] FOX 4 Dallas, “Aviation expert says outdated infrastructure could cause more airport ground stops,” September 22, 2025. https://www.fox4news.com/news/dfw-love-field-ground-stop-experts

    [43] CBS Texas, “Outdated infrastructure is to blame for hours long ground stop at Dallas airports, FAA says,” September 20, 2025. https://www.cbsnews.com/amp/texas/news/outdated-infrastructure-faa-dallas-airport-ground-stops

    [44] Reuters via ARY News, “Dallas airports return to normal after over 2,000 flights disruption,” September 22, 2025. https://arynews.tv/dallas-airports-return-to-normal-after-over-2000-flights-disruption

    [45] FOX 9, “FAA head says billions more needed to modernize air traffic control,” September 16, 2026. https://www.fox9.com/news/faa-head-billions-modernize-air-traffic-control

    [46] The Air Current, “U.S. ends halt on flight departures after nationwide air traffic communications outage,” November 13, 2026. https://theaircurrent.com/feed/dispatches/u-s-halts-flight-departures-nationwide-on-air-traffic-communications-outage/

    [47] Kaia Hubbard, CBS News, “Ground stop proves ‘we have to get off old network,’ Transportation Secretary Sean Duffy says,” November 15, 2026. https://www.cbsnews.com/news/ground-stop-faa-telecom-network-sean-duffy-face-the-nation/

    [48] DISA, Enterprise Voice, Video, and Messaging Policy Security Requirements Guide, Version 1, Release 4, finding V-259917, December 23, 2025 (STIG Viewer). https://www.stigviewer.com/stigs/enterprise_voice_video_and_messaging_policy_security_requirements_guide/2025-05-29/finding/V-259917

    [49] FAA, Order JO 6000.36B, National Airspace System Service Survivability (cancels Order 6000.36A, Communications Diversity, 1995). https://www.faa.gov/regulations_policies/orders_notices/index.cfm/go/document.information/documentID/1037633

    [50] FAA, Order JO 6500.28, Radio Communications Requirements for Air Traffic Control Facilities. https://www.faa.gov/documentLibrary/media/Order/FAA_Order_JO_6500.28.pdf

    [51] 47 CFR § 4.9, Outage reporting requirements, threshold criteria (Legal Information Institute). https://www.law.cornell.edu/cfr/text/47/4.9

    [52] 47 CFR § 4.5, Definitions of outage, special offices and facilities, 911 special facilities, and 988 special facilities. https://www.law.cornell.edu/cfr/text/47/4.5

    [53] 47 CFR § 4.2, Availability of reports filed under this part (Legal Information Institute). https://www.law.cornell.edu/cfr/text/47/4.2

    [54] MPR News, “Flights resume at Minneapolis-St. Paul International Airport after ‘telecommunication issues’,” August 6, 2026. https://www.mprnews.org/story/2026/08/06/flights-halted-at-minneapolisst-paul-international-airport

    [55] MPR News, “Outage at Minneapolis data center behind last week’s telecommunications disruptions,” August 10, 2026. https://www.mprnews.org/story/2026/08/10/how-an-outage-at-unassuming-building-in-minneapolis-sent-waves-through-the-midwest

    [56] FCC, Facilitating Implementation of Next Generation 911 Services; Improving 911 Reliability, Second Report and Order, FCC 26-39, adopted June 25, 2026. https://docs.fcc.gov/public/attachments/FCC-26-39A1.pdf

    [57] FCC Public Safety and Homeland Security Bureau, “Public Safety and Homeland Security Bureau Encourages Communications Service Providers to Implement Important Network Reliability Practices,” Public Notice DA 20-1406, November 25, 2020. https://docs.fcc.gov/public/attachments/DA-20-1406A1.pdf

    [58] Data Center Frontier, “Nashville Explosion: AT&T Repairs Damage to its Building and Network,” December 2020. https://www.datacenterfrontier.com/featured/article/11428484/nashville-explosion-at038t-repairs-damage-to-its-building-and-network

    [59] CISA, “Telecommunications Service Priority: Priority for Emergency Communications.” https://www.cisa.gov/sites/default/files/publications/TSP%20Priority%20for%20Emergency%20Communications.pdf

    [60] NTIA, DOT\FAA 1755-1780 (Rev. 1), Freq-Geo Transition Timeline, FTI Last Mile Microwave for the Elwood ARSR, approved October 3, 2018. https://www.ntia.gov/sites/default/files/publications/releasable_dot_faa_1755-1780_rev-1_sufficient_10_26_2018_0.pdf

    [61] David Shepardson, Reuters, “US FAA not considering replacing L3Harris, Verizon contracts with Musk’s Starlink,” July 2025 (via Yahoo News). https://www.yahoo.com/news/us-faa-not-considering-replacing-161335458.html

    [62] RISKS Digest, Volume 10, Issue 62, “Telephone cable cut eliminates O’Hare tower communications,” November 19, 1990 (reprinting The New York Times, October 16, 1990). https://catless.ncl.ac.uk/Risks/10/62

    [63] RISKS Digest, Volume 10, Issue 75, “NY area fiber-optic telephone cable severed,” January 5, 1991. https://catless.ncl.ac.uk/Risks/10/75

    [64] Associated Press via Fox News, “FAA Outage Exposes Odd Computing Practices at Agency,” August 2008 (the page shows January 13, 2015). https://www.foxnews.com/story/faa-outage-exposes-odd-computing-practices-at-agency.print

    [65] Air University, “Air National Guard restores air traffic control in Puerto Rico,” October 2017. https://airuniversity.af.edu/News/Display/Article/1331249/air-national-guard-restores-air-traffic-control-in-puerto-rico

    [66] NOAA, “Federal collaboration yields radar coverage for Puerto Rico, USVI in wake of Hurricane Maria,” October 2017. https://www.noaa.gov/media-release/federal-collaboration-yields-radar-coverage-for-puerto-rico-usvi-in-wake-of-hurricane

    [67] CNN via ABC 17, “Nashville explosion causes AT&T outages and disrupts flight,” December 25, 2020. https://abc17news.com/news/national-world/2020/12/25/nashville-explosion-causes-att-outages-and-disrupts-flight/

    [68] Colorado Public Radio, “FAA investigating outage that interfered with communication between pilots and Denver air traffic control,” May 15, 2025. https://www.cpr.org/2025/05/15/faa-investigation-dia-air-traffic-control-outage/

    [69] Associated Press via WUSF, “Cut fiber line near Jacksonville briefly disrupts air traffic controllers around Southeast,” June 20, 2025. https://wusf.org/text/local-state/2025-06-20/cut-fiber-line-near-jacksonville-briefly-disrupts-air-traffic-controllers-around-southeast

    [70] AVweb, “Severed Fiber Line Disrupts Raleigh-Durham Flights,” October 13, 2026. https://avweb.com/aviation-news/severed-fiber-line-disrupts-raleigh-flight/

    [71] AVweb, report on fiber damage and the ground stop for Houston Intercontinental, October 14, 2026. https://www.avweb.com/aviation-news/severed-fiber-line-disrupts-houston-flight/