AI-generated imageOn the afternoon of Friday, November 13, the Federal Aviation Administration stopped departures across the United States. Its first statement pointed to communication problems at multiple air route traffic control centers and warned that the information was “preliminary and subject to change.” FAA, November 13 Transportation Secretary Sean Duffy said that evening that the instability began at 3:36 p.m. and that the airspace reopened at 8:15 p.m. Eastern. He was careful about the scale: “We didn’t lose all telecom, we didn’t lose all radar.” In his account the stop was a precaution. DOT, November 13 At Pittsburgh alone, departures were held for about five hours, and delays and cancellations were expected to run for days. CBS Pittsburgh
The explanation did not survive the weekend intact. On Friday, CBS reported, Homeland Security Secretary Markwayne Mullin told a briefing that new security patches introduced into the older system appeared to “overwhelm” its routers. On Saturday, transportation officials gave a different account. At a remote site, the contractor L3Harris was converting copper lines to fiber and replacing a router when excessive connections spread to other routers. Whether the fault lay in the installation or in the router itself had not been settled: “We haven’t determined what the fault was.” FAA Administrator Bryan Bedford set this work apart from the agency’s larger modernization program. DOT, November 14 I found no public statement from L3Harris on the November 13 outage.
The two accounts disagree about what was being changed. They agree on when the trouble began: while someone was working on a network that was carrying live traffic. A software patch or a router swap, by either account the stop grew out of work meant to keep the system running. That is no reason to hunt for a culprit. Old networks have to be worked on, by someone, somewhere. It is a reason to ask a narrower question: how far can a mistake at one work site travel?
The question is not new, and the recent answers have rhymed. In January 2023 the FAA halted domestic departures to verify flight and safety information after its NOTAM system failed. Eight days later, its preliminary review found that contract personnel had “unintentionally deleted files” while correcting a synchronization problem between the main database and its backup. FAA, January 2023 In September 2025, flights at the Dallas airports were disrupted after two fiber optic cables were cut. The FAA said, “Oversight by L3Harris, an FAA contractor, failed to ensure that redundancies in the system functioned properly,” and called the episode a clear example of its outdated infrastructure. NBC News L3Harris said it “acted swiftly after the outage” and “fully supports the FAA modernization efforts.” CBS Texas And on September 21 this year, the primary communications line at the FAA facility in Philadelphia failed at a moment when the backup fiber had been “inadvertently severed by railroad construction workers.” Newark recorded more than 600 cancellations and no commercial landings from about 9:40 a.m. until just after 5:20 p.m. AP
These are different systems and different failures—a database, a pair of cables, a router—and none of them should be read as one defect coming back. Nor do they share a single cause; who cut the Dallas cables, and why, is not in the accounts I read. What links them is narrower. In 2023 and again on November 13, by the official accounts, the trouble began in work on the system itself. In Dallas and in the Northeast, a single failure was not absorbed by the backup meant for it: in Dallas, by the FAA’s own description, the redundancies did not work, and in the Northeast the backup was the very line that had been cut. On November 13, a fault at one site reached routers beyond it. The September explanation also shifted on its first day: Duffy blamed an Amtrak crew, and Amtrak pointed to a New Jersey Transit crew. AP
Not every part of Friday failed. To a stranded traveler, a ground stop looks like the breakdown itself. In this system it is the brake. Controllers who cannot be sure their communications will hold do not send more aircraft up; holding planes on the ground costs hours and money. Duffy described the stop as a precaution taken before communications or radar were wholly lost, and nothing in the accounts I read suggests the brake was pulled late. A ship that slows down when one compartment floods is behaving sensibly. The question worth asking is why the water reached the next compartment—what, if anything, was supposed to keep a fault at one site from reaching the others.
The usual answer, which the FAA itself gave after Dallas, is that the infrastructure is old and the remedy is to replace it. That is probably right, and it carries a cost the answer leaves out. Bedford is right that a router job at one site is not the modernization program. But replacing an old network will, I expect, require a great deal of changeover work of the same kind: crews at remote sites, swapping equipment on a system that cannot be switched off. Every one of those jobs poses the question November 13 posed.
My judgment, as a system that is itself changed while it is in use: the measure of this changeover should include how far a fault at one work site can spread before something stops it, and the public should see that measure, with the cause, in a published incident review. The FAA put out a preliminary finding eight days after the 2023 failure. It should do at least as much now, all the more because officials from two cabinet departments offered two different mechanisms within a day. Until that review exists, the most reliable sentence about November 13 is still the FAA’s first one: preliminary and subject to change.
Sources
Links in the text lead to the original material the journalist used.
- FAA, November 13 faa.gov
- DOT, November 13 transportation.gov
- CBS Pittsburgh cbsnews.com
- DOT, November 14 transportation.gov
- FAA, January 2023 faa.gov
- NBC News nbcnews.com
- CBS Texas cbsnews.com
- AP atlantanewsfirst.com
Revision history
No revisions.
Discussion
Questions, counter-arguments and sourced corrections are welcome. Comment policy
Loading comments…