Just before 4 pm Eastern on Friday 13 November 2026 the FAA stopped every departure in the United States. Flights resumed at 8:15 pm, after the longest halt to departures since 11 September 2001. It was not a cyberattack, officials said repeatedly. The next morning Transportation Secretary Sean Duffy said the preliminary investigation pointed to a router brought into service at a remote FAA site by a contractor moving the network from copper to fiber.

The network is the FAA Telecommunications Infrastructure, FTI, and the contractor is L3Harris. Seventeen years ago almost to the week, on 19 November 2009, the same network, run by L3Harris's predecessor Harris, failed across the country during a migration to fiber, because of one router. That time there is a finished investigation. This time there is not, yet. This piece puts the two side by side.

[INFO]
Status: the 2026 cause is preliminary. Duffy said it was unclear whether it was installer error or a faulty router, and FAA Administrator Bryan Bedford said a root cause analysis would follow. The 2009 account is from the DOT Inspector General's letter CC-2010-012 (17 June 2010). Statements about 2026 are from the 14 November briefing as reported by CNBC and CNN, and from CBS News's government sources. Read on 15 November.
◈ interactive artifact
One Router, Every Centre
Bring the router back into service in 2009 and in 2026, watch what the network does, and decide where to look and whether to stop the departures.

//What Happened on 13 November

The FAA's first statement cited communications issues "impacting multiple Air Route Traffic Control Centers", the 22 centres that handle aircraft at cruising altitude. At the briefing the next day Bedford said as many as 17 of the 22 were affected, but that none "actually went dark". What failed was the stability of the information feeds into them. For about 45 minutes officials did not know why, and Bedford said the FAA was "unwilling to take any risk" that a centre would lose contact with aircraft. So everything on the ground stayed there, and the roughly 3,100 US-bound aircraft already in the air (Flightradar24, via CBS) landed.

FlightAware counted 9,159 US cancellations that Friday, the second-worst day of the year after a January snowstorm, and thousands more on Saturday while crews and aircraft were out of position.

//The Router

Duffy's account: as part of L3Harris's copper-to-fiber work at "at least" one remote location, a router began flooding other air traffic control routers with "over 1,000 connection attempts roughly every five seconds", disrupting communications and radar feeds. Investigators, he said, were trying to establish whether it was "user error, installator error from L3 Harris, or did they install a faulty router". L3Harris said it was supporting the investigation and that it had completed more than 70% of the copper-to-fiber switch.

Other accounts do not yet line up. CBS News's senior government sources said the problem was limited to Cisco 600 series routers made between 2001 and 2006, and that the faulty equipment had been isolated. Other CBS sources said a software update to the network caused a reboot. Homeland Security Secretary Markwayne Mullin said new security patches appeared to have overwhelmed the older system's routers. The Wall Street Journal placed the site in southern Arkansas; this site could not read that report. Treat all of it as preliminary.

//Then: 19 November 2009

The Inspector General's letter is unusually exact. Harris was moving FTI's IP services from a legacy network to a new fiber network, using temporary routers called Logical Transition Bridges at 26 core sites to keep the two apart. On 21 October 2009 a temporary router at the Los Angeles centre was configured with an error that let the network "believe" traffic could take a route it should not. When the link to Salt Lake City came back up, the correct routes resumed and hid the mistake, which lay dormant.

On 19 November a Harris engineer took that link down to put a new router in service at Los Angeles. The route map was copied from the old router, error included, and the engineer did not append a needed "no" statement before bringing the link back. The network began sending all its traffic through Los Angeles to Salt Lake City over a link that could not carry it. About 75% of routes on the fiber network were blocked for five hours. Every en route centre and the FAA Command Center lost the flight-plan, traffic-management and defence surveillance data they relied on. According to Harris, 129 facilities were hit instead of the 21 that should have been. 820 flights were delayed.

The five hours were the second failure. An alert that fires when a router's CPU stays above 60% for 10 minutes never fired, because a filter meant to silence repeat alarms from one router at the FAA Technical Center had been applied to every router on the network. Engineers then chased the wrong problems: backing out the latest changes at Los Angeles and Herndon, and sending a technician to reset a Los Angeles router they could not reach. Their spot checks of core routers skipped Salt Lake City, where the overload was. Service came back when someone took a core router there offline.

//What Is the Same, and What Is Not

The same: one contractor-run network carrying the data every centre depends on; a migration to fiber, done live; and, if Duffy's account holds, one router at one site whose behaviour spread across the country. In 2010 the Inspector General warned that "an inherent risk of critical outages remains" as more services moved onto the new network, and that the FAA's oversight of the contractor was reactive. The independent review panel's recommendations included automated tools for router configuration changes with independent verification, checklists designed with human error in mind, end-to-end awareness of the network, and a model of FTI to test changes on before making them.

Not the same: in 2009 the outage hit flight plans and traffic management data and grounded hundreds of flights, but there was no nationwide ground stop. In 2026 the instability reached the centres' communication and radar feeds, and the FAA chose a full ground stop. The mechanism may differ too: 2009 was a routing error that sent everything down one link; 2026, on the account so far, is a device generating traffic until its neighbours could not cope. Until the root cause analysis is out, the honest comparison is between a finished investigation and a preliminary statement.

The last nationwide ground stop before this one, on 11 January 2023, was a different system: the FAA later said contract personnel had "unintentionally deleted files while working to correct synchronization between the live primary database and a backup database" of the NOTAM system. Three nationwide failures, each during contractor maintenance, and none found to be malicious - the 2026 one on preliminary evidence. Like the CrowdStrike outage of July 2024, they did more damage than most attacks without anyone attacking.

//What This Says to Anyone Running a Network

The 2010 panel's list reads like a change-management checklist, because it is one. Changes to routing go through a tool that checks them, not a copy and paste. Anything copied from a transition device is reviewed as new configuration. A link coming back up after maintenance is watched, by someone, with the alarms confirmed working, rather than assumed fine. Alarm filters are scoped to the device they were written for and audited. And there is a way to tell, in minutes, which box is generating the traffic, which is the 45 minutes the FAA did not have.

[IOC]
Network: FAA Telecommunications Infrastructure (FTI), run by L3Harris (Harris Corporation in 2009)
13 Nov 2026: ground stop just before 16:00 to 20:15 ET; up to 17 of 22 en route centres affected, none dark; preliminary cause a router at a remote site during copper-to-fiber work (Duffy); root cause pending
19 Nov 2009: router configuration error during migration to fiber; ~75% of routes blocked for 5 hours; 129 facilities; 820 flights delayed (DOT OIG)
11 Jan 2023: NOTAM database files deleted by contract personnel (FAA); separate system

//Sourcing and Uncertainty

The 2009 account, figures and quotes are from DOT OIG letter CC-2010-012 and its enclosure listing the review panel's recommendations. For 2026, the ground stop times, Duffy's quotes, Bedford's "unwilling to take any risk" and 45 minutes, and L3Harris's statement are from CNBC; the 17 of 22 centres and "actually went dark" are from CNN; the Cisco routers, the conflicting software-update account, Mullin, the 3,100 aircraft and FlightAware's 9,159 are from CBS. The 2023 quote is the FAA statement of 19 January 2023 as reported by CNBC.

What is unknown. Whether the router was misconfigured or faulty; whether the software-update account and the router account describe the same event; which 17 centres were affected; why one router's traffic could destabilise feeds across the country; and what the FAA's root cause analysis will say. This article will be updated when it is published.