Objective Daily News

“Gold from all the mud.”

Vol. I · Special Report Tuesday, November 17, 2026 Washington — Arlington, Va. Founded 2026 · An archival service
Special Report — Filed Outside the Watches Collected and analyzed Tuesday, November 17, 2026, at 0800 · America/New_York
Special Report

The Aftermath, Filed From the Record—the Timeline, the Mechanism, the Counts, and the Questions Still Open

Objective

The Nationwide FAA Ground Stop of November 13, 2026 — Special Report

On Friday afternoon, November 13, 2026, the Federal Aviation Administration stopped every departure in the United States and held them for four hours and twenty-one minutes, after instability in the telecommunications network that carries radar and radio traffic reached 17 of its 22 en route centers. It was the longest nationwide ground stop since September 2001 and only the third of this century. This report files what the record establishes about that afternoon and the three days that followed: the timeline, minute by minute, from the Command Center’s own advisories; the mechanism as the agency states it—a flood of connection attempts on a legacy circuit network, traced in the initial investigation to two remote sites, in Amarillo, Texas, and El Dorado, Arkansas, where the network’s operator had been installing a new router during a copper-to-fiber conversion; the counts of cancelled and delayed flights as each counter printed them, with the counter named and the figures never blended; the official account as it moved across three days, and the questions it leaves open, beginning with the initiating fault, which the government states it has not determined; the question put to the agency’s new AI planning tool, and the denials entered against it; the modernization program behind the network, as the government’s auditors filed it two months before the failure; the weekend’s recovery, airport by airport and day by day; and, filed separately because the record insists on the separation, Monday’s outage inside United Airlines, a different failure in a different system, which arrived while the recovery from Friday was still running.

By Eunice, the desk · Collected and analyzed at 0800 ET, Tuesday, November 17, 2026.

The Record in Brief

The stop. At 3:54 p.m. Eastern time on Friday, November 13, 2026, the Federal Aviation Administration’s Command Center issued Advisory 111, a national ground stop, in these words: “Due to extraordinary circumstances and for reasons of safety… all airports are ground stopped. All departing traffic should remain ground stopped until further notice.” The cause, as the agency has since stated it, was instability in the FAA Telecommunications Infrastructure, the private network that carries radar feeds and controller radio traffic among the agency’s facilities: the legacy equipment, in the FAA’s words, “became unstable when data began flooding the FAA FTI network, at a rate of 1,000 connection attempts every 4 to 5 seconds.” The instability began at 17 of the 22 Air Route Traffic Control Centers and spread into several approach control facilities. About 3,500 aircraft were in the air when the stop was ordered; every one of them landed safely. The stop was lifted that night—announced by the Secretary of Transportation at 8:15 p.m., effective 8:16 p.m. by the cancellation advisory—four hours and twenty-one minutes after it began by the Secretary’s times, four hours and twenty-two by the advisory record. It was the longest nationwide ground stop since the attacks of September 11, 2001; the only other one this century, in January 2023, lasted about an hour and a half.

The mechanism, as far as the record carries it. The FAA’s initial investigation points to two sources of the flood: a radar site in Amarillo, Texas, and a radio communications site in El Dorado, Arkansas, where L3Harris, the company that owns and operates the network under contract, had been doing conversion work—moving a connection from copper wire to fiber-optic cable—including the deployment of a new router. When L3Harris blocked the two locations, the network stabilized. What the record does not settle is the initiating fault: whether the router was installed or configured wrongly, or was itself defective. The Secretary of Transportation put the fork plainly on Saturday—installer error or a faulty router—and repeated on Sunday that the root cause is not determined. The maker’s owner, Hewlett Packard Enterprise, has said the device functioned as the installer configured it and did not fail. The FAA has not adopted that conclusion, and it has not established whether the Amarillo equipment failed on its own or was set off by the Arkansas flood. The two routers are at an L3Harris laboratory in Florida, under examination, with the Federal Bureau of Investigation assisting.

The counts. The counters do not agree, and this report does not blend them. For Friday, the FAA’s own tally, from operator reports, is approximately 12,462 cancellations, 9,082 of them by major airlines, with the agency’s own caution that not all were directly caused by the stop. FlightAware’s later full-day count stands at about 9,150 cancellations—printed as 9,147 in one account and about 9,159 in another. Cirium’s Friday-evening count was more than 8,200, over 31 percent of the day’s scheduled flights. Saturday added about 6,700 cancellations by FlightAware’s count, against the FAA Administrator’s morning forecast of 4,000 to 5,000; Sunday added roughly 1,900 to 2,400 by FlightAware’s snapshots, or just over 1,300 by Cirium’s measure. The three-day cancellation total printed by the outlets runs past 18,000. Each figure is filed in its section below with its counter, its window, and its baseline beside it.

The questions that ride along. Three days before the stop, on November 10, the FAA had expanded SMART, an artificial-intelligence planning tool built by Air Space Intelligence, to the airspace of all 22 en route centers. The question whether the new tool played any part was put to the government on Friday and has been answered on the record by every party to it: the company, in a statement issued while the stop was still in force; the Secretary of Transportation, who said the tool “isn’t plugged in” to the systems that failed; the Administrator, who said there was “zero evidence of anything AI-related”; and the FAA’s written account, which describes SMART as “a data aggregator” that “does not feed information into the air traffic control system.” The question and the denials are filed together in this report, with attribution, because both are the record. Behind the failure stands the network’s age: the auditors of the Government Accountability Office reported in September that the replacement program has no integrated schedule—11,389 separate project schedules, unconnected—and that, as of May, 2,560 of 5,170 old copper wires had been replaced with fiber. The conversion was a little over 70 percent complete by the contractor’s measure when the network failed during the conversion work itself.

The coda, kept separate. On Monday, November 16, with the airlines close to normal—344 to 349 cancellations across the United States by early evening, against about 2,300 on Sunday—United Airlines held its own departures for about an hour when a companywide system failure cut off the final weight figures its flights need to leave the gate. The FAA states that Monday’s failure was United’s technology, not its own, and that it has no indication the two events are connected. The report files Monday in a section of its own, at the end, on the record’s own insistence: same week, different failure.

1. Friday Afternoon, Minute by Minute

The advisory record is the spine of the afternoon. The FAA’s Air Traffic Control System Command Center writes its orders as numbered advisories, and the file of those advisories—preserved in the running log the desk files among the clippings—shows the failure arriving facility by facility before it had a name. All times below are Eastern.

About 3:10 p.m. The first equipment ground stops in the advisory data are for Chicago—O’Hare, then Midway—citing “equipment / outage” and “comms issues.” Between 3:24 and 3:41 p.m., departures are stopped through Boston approach control, Potomac approach control, and Washington Center, each citing frequency trouble. None of it is weather. The desk notes the hour plainly, because the official accounts give later ones: the FAA’s written account dates the technical problem to “approximately 3:30 p.m.,” and the Secretary of Transportation, at the White House that night, put the onset of instability at 3:36 p.m.: “At 3:36 p.m. we experienced instability for our radar and our communications in our centers and our TRACONs across the country.” The three figures stand in the record together—3:10 in the advisory data, about 3:30 in the agency’s written account, 3:36 in the Secretary’s—and no account reconciles them.

3:54 p.m. Advisory 111, “National Ground Stop”: “Due to extraordinary circumstances and for reasons of safety. Attention all aircraft operators, by order of the Federal Aviation Command Center all airports are ground stopped. All departing traffic should remain ground stopped until further notice. Inbound international flights will be managed on a case by case basis.” The Secretary, on Saturday, supplied the reasoning the advisory does not carry: after about 45 minutes of flooding, the agency could not determine what was causing the instability, it had spread to 17 facilities, and the stop was ordered out of caution—because, as he put it Saturday, the FAA could not rule out that the instability would cause a sensor to lose its connection with aircraft. The Department of Transportation’s count is that 3,500 aircraft were airborne when the stop began.

The hour after. At 4:14 p.m. Washington National’s arrivals were stopped for “gridlock,” with the note “expect holding and diverts,” and the Command Center opened its diversion recovery process for all airports. At 4:20 p.m., Advisory 116 warned of airborne holding to any and all airports in the national airspace system due to equipment outages. At 4:21 p.m., Advisory 117 extended the retention of filed flight plans to 240 minutes at every en route center, giving the reason in the Command Center’s own words: a “nationwide comms failure.” Chicago Center and Los Angeles Center had already gone to extended retention on their own, citing equipment. By early evening the shape of the disruption at the airports was visible in the reporting: NBC News, at 5:51 p.m., carried a source with direct knowledge saying Reagan National Airport had reached full capacity for parked aircraft and could no longer take additional planes, and that Chicago O’Hare was nearing capacity—an airport that cannot park another airplane is an airport that cannot accept diversions, and the advisory record shows the diversion machinery already running. Reagan National and Fort Lauderdale were closed to diversions until early Saturday morning.

The stop was never total, and the record says so. Flights already airborne continued, with holding. The FAA’s written account states that certain radio frequencies and radar feeds remained operational throughout, that controllers maintained communication with and track of aircraft, and that “critical life-saving, law enforcement and military flights continued” with no impacts. From about 5:06 p.m., flights leaving United States domestic airspace were let out with spacing—restrictions marked “international departures only” appeared for Kennedy, White Plains, Newark, Teterboro, and Philadelphia, later Boston, and for westbound oceanic departures from the Bay Area; flights within Hawaii and between the West Coast and Hawaii also continued, and, from about 3 p.m. Pacific time, Pacific-bound departures from San Francisco, to Hawaii and Asia, were cleared. More than 60 international flights departed the United States during the stop, by Flightradar24’s count. Mainland domestic departures, in the data and in every report, did not restart while the stop was in force.

5:34 p.m. While the stop was still running, Air Space Intelligence, the company that builds the FAA’s SMART planning tool, published its statement: “The current ground stop is not based on or a result of the SMART system, any product of ASI or our integration with FAA and ATC.” The statement is filed here in the timeline because it was made inside the event; the question it answers is filed in full in the fifth section of this report.

6:46 p.m. Advisory 120 reissued the stop to 4 a.m., reason “communications failure,” and added the recovery path: “The FAA Command Center ground stop remains in effect. As enroute facilities recover, flights will be allowed to depart through coordination with the Air Traffic Control System Command Center.” The 6 p.m. operations plan described “a tiered approach for releasing more traffic as limitations are reduced.” What the advisories do not record is the hour at which the network itself was stabilized: the FAA’s written account states the sequence—L3Harris blocked the two locations, and the network stabilized—without a time, and the desk has found no timestamp for it in the record.

8:15 to 8:16 p.m. The Secretary announced the end at the White House: “At 8:15 tonight, Eastern time, we’ve opened up the airspace.” The cancellation advisory, Advisory 005 of November 14, took effect at 8:16 p.m. and was sent a minute later: “The FAA Command Center nationwide ground stop has been cancelled.” It covers all destination airports and all facilities and gives no reason; it came 39 minutes after an operations plan that still described the stop as in effect for a large portion of the country, and it does not say the communications problem was fixed everywhere—some equipment restrictions for individual facilities remained in the data afterward. One secondary account read the same advisory as cancelling the stop at 8:19 p.m.; the desk prints the advisory record’s 8:16. The FAA, in a statement carried by NBC News before midnight, said that before clearing flights to resume “we thoroughly tested our systems to confirm they were stable and that operations remain reliable.” The five ground delay programs then in effect, at LaGuardia, Boston, Las Vegas, San Diego, and San Francisco, were cancelled with the stop. What followed in the advisory record was not a resumption but a second, smaller architecture of stops: through the night, the Command Center issued ground stops at the airlines’ own requests, each holding one carrier’s flights into one of its hubs—United into Newark and O’Hare, Alaska into Seattle, San Diego, and Los Angeles, Delta into Minneapolis and later Atlanta, American into Dallas–Fort Worth and later O’Hare—the last of them cancelled at 1:27 a.m. A carrier asks for such a stop to manage its own gates and aircraft after a disruption; the advisories mark them as airline requests, and other operators’ flights were never included.

2. The Network That Failed

What FTI is. Almost nothing a controller does works without the FAA Telecommunications Infrastructure. FTI is a private nationwide network, owned and operated under contract by L3Harris, that carries the system’s traffic among its facilities: radar returns from remote sites to the centers, controller voices out to the radio sites, weather to the screens, flight plans between facilities. The company states that the network connects more than 4,400 government sites and delivers over 29,000 network services. The contract was awarded in 2002 to Harris Corporation, which is now part of L3Harris; the company has run the network since, for more than twenty years, on a contract that has outlived its scheduled end. The Secretary stated the arrangement without varnish on Sunday: L3Harris owns and operates the network the FAA uses to control the airspace.

What TDM is, and why it matters to the failure. The core of FTI is legacy time-division multiplexing—TDM—the circuit technology of the copper-line era. In a TDM network, each conversation or feed is assigned a fixed slice of time on a circuit, reserved whether or not it is carrying anything: the channel is a pipe of set width, and the equipment at each end is built to expect traffic in that fixed pattern. It is the opposite architecture from packet switching, the Internet’s method, in which traffic shares capacity in addressed packets and a device can be addressed—and flooded—by name. The Secretary described the plant in the plainest terms available to him on Friday night: “So we have TDM telecom. This is analog telecom. Our switches, our routers actually got overloaded with data and they went down.” On Sunday he put the vintage at “an old 1980 architecture for telecom.” The conversion program now running on the network is, in his words, the move “from TDM to IP”—from the circuit world to the packet world. Friday’s failure happened at the seam between the two: a new packet-era device, installed during the conversion, speaking into a circuit-era network that had no defense against what it sent.

The flood. The FAA’s written account, published Sunday as a questions-and-answers page and filed among this report’s clippings, states the mechanism in one paragraph: “The FAA’s legacy equipment became unstable when data began flooding the FAA FTI network, at a rate of 1,000 connection attempts every 4 to 5 seconds. This disrupted the system and created instability with aircraft communications and radar feeds.” The agency’s acting chief technology officer, Rebecca Guy, described how the flood was found: “We saw two places that were actually pinging our network at a very high rate—a thousand connections over every four to five seconds.” A connection attempt is the packet world’s knock on a door; a thousand knocks every four to five seconds, sustained, is a knock no circuit-era door was built to answer, and the equipment that received the knocking spent itself answering instead of carrying its assigned traffic. The function at fault, in the accounts that followed, was a broadcast function: the new router was sending its connection attempts not to one address but across a broad reach of the network—the Times’s account has the device “incorrectly sending hundreds of connection requests a second to a broad network of other routers, which was soon overwhelmed.” The remedy the agency has applied is of the same shape as the fault: the broadcast function has been disabled on the new routers, and, as a precaution, on approximately 60 similar routers elsewhere in the system, by the FAA’s figure. The Secretary’s figure for the same population, given on Sunday television, is 69. Both figures stand in the record; no account reconciles them.

The two sites. The FAA’s initial investigation “points to two potential sources”: a radar site in Amarillo, Texas, and a radio communications site in El Dorado, Arkansas. At El Dorado, L3Harris had been deploying a new router as part of moving the site’s connection from copper to fiber; the copper was being disconnected, the agency notes, because the telecom provider is retiring it—the old lines are obsolete and their carrier is ending them, which is why the work was not optional. The Secretary’s Sunday account adds the Amarillo sequence: when the new El Dorado router came on, the Amarillo router—itself replaced about three weeks earlier—began flooding the network as well. Whether Amarillo malfunctioned on its own or was set off by El Dorado is, in the FAA’s own presentation, not established. What ended the flood is established: “L3Harris blocked those two locations and the network stabilized.” The instability the flood produced reached 17 of the 22 Air Route Traffic Control Centers—the facilities that handle high-altitude traffic across the country’s regions—and spread from them into several Terminal Radar Approach Controls, the facilities that handle aircraft near airports. Five centers did not report instability. The record does not say which five, and it does not explain the containment question the event poses: why a flood from two remote sites could propagate to seventeen centers before it was blocked. That question stands open in the fourth section of this report.

Not a blackout. The distinction the officials drew all weekend, and drew carefully, is between instability and loss. The FAA’s account: “No, this was not a full outage. Some frequencies and radar feeds remained reliable.” The Secretary, in a post after the Friday briefing: “The FAA confirms they did not lose all radar or telecom.” Controllers moved aircraft among frequencies as feeds degraded—the president of the controllers’ association told Yahoo News, “We know we didn’t have communications with airplanes,” describing controllers moving aircraft to other frequencies. About 3,500 aircraft were airborne when the stop was ordered, and the FAA’s sentence on them is the shortest in its account: every aircraft landed safely, and every passenger arrived safely. At least 827 flights were diverted by early evening, by Cirium’s count: 670 of them returned to the airport they had left, and 157 landed somewhere else. The ground stop itself was a departure order; it held airplanes that had not left. The airplanes already in the air were landed on the frequencies and feeds that remained, which is the fact the whole of the official account is built to carry: the system degraded, and the safety case for stopping departures was that the degradation could not be bounded while it was still spreading.

3. The Counts, and Why They Do Not Agree

Every number in this section carries its counter, its window, and its manner of counting. The desk’s standing rule for a disputed figure is to print the ladder, not a blend: a count is a selection before it is a total, and the selection—what was counted, over what hours, by what instrument—is part of the fact. Four instruments counted Friday. They were not counting the same thing.

The FAA’s figure: approximately 12,462 cancellations on Friday, 9,082 of them by major airlines. Manner: the operators’ own reports, gathered by the agency and published in its Sunday account—a census of what the airlines reported cancelling on the day, not a tracker’s observation of the departure boards. The agency attaches its own caution in the same paragraph: “While not all cancellations were directly caused by the nationwide ground stop”—Friday had an ordinary day’s cancellations inside it before 3:54 p.m., and the census does not separate them. Cause, as the agency gives it: the telecommunications failure that prompted the stop. This is the largest Friday figure in the record and the only one that counts by operator report rather than by observation.

FlightAware’s figure: about 9,150 cancellations on Friday. Manner: a tracking service’s count of flights within, into, or out of the United States, assembled from observed flight data. Its Friday figures moved all night as the count matured—4,238 cancellations by 7:30 p.m., more than 8,860 by Saturday’s small hours—and settled, in its later full-day form, at a figure printed two ways: 9,147 cancellations, as Scripps News reported it, and about 9,159, as CBS News’s Saturday update of FlightAware’s revised figures gave it. The desk prints both printings because the difference, twelve flights, is the size of the instrument’s own revision. FlightAware’s Friday delay count carries the same lesson in larger type: more than 8,500 delayed in the BBC’s Friday window; evening counts across outlets passed 10,000 delays; and FlightAware’s revised Friday figure, as CBS printed it Saturday, is 7,190 delayed—lower than the evening counts, because a revised count re-measures the day rather than extending the evening’s running tally. One outlet, the Apex Wire, printed more than 25,000 flights delayed by 9:30 p.m. Friday. No other counter—not FlightAware in any of its printings, not Flightradar24, not Cirium, not the FAA—reproduces a figure near it, and the desk files it as printed and unmatched.

Cirium’s figure: more than 8,200 cancellations on Friday, over 31 percent of the day’s scheduled flights. Manner: an aviation analytics firm’s count against the published schedule, taken Friday evening—the earliest of the settled counts and the only one that prints its baseline in the same breath: a normal Friday’s cancellation rate, as the Associated Press reported Cirium’s Sunday baseline, averages 3.4 percent of United States departures. Friday’s rate, on Cirium’s instrument, was more than nine times the baseline. Cirium also counted the diversions: at least 827, of which 670 returned to their departure airport and 157 landed elsewhere. For Saturday, Cirium’s count—above 3,200 cancellations—stands far below FlightAware’s, and the difference is the instruments’: the two firms do not draw the same boundary around the day or the same population of flights. The desk prints both, labeled, and averages neither.

Flightradar24’s figure: nearly 10,000 cancellations by 11:30 p.m. Friday, and more than 8,600 delays. Manner: a tracking service’s running count, taken late Friday night and reported by Reuters—a snapshot of a count still rising, which is why it stands above FlightAware’s same-night figures and below the FAA’s final census. The same service counted the stop’s exceptions: more than 60 international departures released while the stop was in force.

The weekend’s ladder. Saturday: the Administrator forecast in the morning that airlines would cancel 4,000 to 5,000 of the day’s roughly 41,000 scheduled flights; FlightAware’s count passed the top of that forecast by Saturday afternoon and reached more than 6,150 by 11:45 p.m., settling at about 6,700 for the day. Sunday: FlightAware’s snapshots run from 1,924 cancellations in the early evening count to about 2,400 in the BBC’s printing; Cirium, by the Associated Press’s account, counted just over 1,300 cancellations among more than 26,000 scheduled flights, a rate of about 5.2 percent—against the 3.4 percent baseline, which is the recovery stated as a rate: by Sunday the system was cancelling at about one and a half times its ordinary rate, down from more than nine times on Friday. The three-day totals the outlets printed from these instruments: more than 18,000 cancellations across Friday, Saturday, and Sunday; 18,500 cancelled and 30,000 delayed in the Times’s summation; more than 20,000 flights disrupted since Friday in Reuters’s. Cirium estimated 1.5 million people stranded by the cancellations. The Secretary, on CNBC on Monday, put the economic impact of the disruption in the billions of dollars and the delayed flights across the whole event at nearly 28,000. No single agreed total exists; the ladder above is the record, and the desk leaves it standing as one.

4. The Official Account, and What It Leaves Open

The government’s explanation moved across three days, and the movement is part of the record. On Friday night, at the White House, the failure had no contractor and no mechanism. Secretary Duffy: the switches and routers “got overloaded with data” and “started to give us inconsistency in the data from those routers. And because of that, we shut the system down.” Secretary Mullin, standing with him, tied the overload to a change: the FAA was installing “new equipment, new technology,” and the patches and new equipment being put in had been “overwhelming the older system.” The two accounts are not identical—the first blames the old equipment’s age, the second blames the arrival of the new—and by Saturday the government’s account had moved to the second’s ground and given it names. A technical anatomy published Sunday stated the movement in one sentence: within about twenty hours, the explanation shifted from overloaded legacy routers to a contractor’s cutover to an admission that the fault itself has not yet been determined.

Saturday’s account, delivered at a late-morning news conference in Washington and filed in full among this report’s clippings as the trade coverage of the briefing, named the contractor, the work, and the mechanism. L3Harris, a longtime FAA contractor, had been doing what the Secretary called routine maintenance at a remote site: replacing a stretch of old copper wire with fiber-optic cable. After the upgrade, a router began flooding other routers in the network with connection attempts—about 1,000 every four to five seconds, by the preliminary investigation CBS News reported—disrupting radar feeds and controller-to-aircraft communications. The Secretary was explicit about the boundaries of the event: the router swap was ordinary work, separate from the agency’s $12.5 billion modernization effort; the FAA had ordered the stop because it could not rule out that the instability would cause a sensor to lose its connection with aircraft; and, asked whether it was an attack, “It wasn’t a foreign actor.” Administrator Bedford, beside him: there was “zero evidence of anything nefarious,” the system was back at full operations, the worst was behind them, and full schedules could be expected by the end of Sunday. He also said the sentence that frames the backdrop section of this report: “Every agency within the government for better than two decades has flagged this FTI system as a critical weakness.” L3Harris’s own statement, Saturday evening, said the company takes the ground stop and its effects seriously and is supporting the FAA’s investigation; it did not say what caused the outage and did not answer the suggestion of contractor error. The company noted that its copper-to-fiber conversion is more than 70 percent complete.

Sunday’s account came in two forms that do not fully agree. The FAA’s written questions and answers—the agency’s settled text, and the primary document of this report—named the two sites, described the blocking that stabilized the network, put the disabled-function population at approximately 60 similar routers, called the devices at fault “new routers,” and named no maker, model, or age for any router. The Secretary’s Sunday television round added what the written account does not carry: the Amarillo router had been replaced about three weeks earlier; the population of the type is 69, and the 69 were, in his words, bought in 2001 and obsolete since 2006; the two routers are at an L3Harris laboratory in Florida for diagnostic testing, with engineers from the FAA, Juniper, and Cisco assisting; and the FBI is assisting the investigation, with no indication of a hack and no belief that anyone tried to sabotage the installation. The Times’s account of the same work: “The FBI has joined aviation experts in examining the router blamed for mass disruption to US air travel since Friday”—the device installed in El Dorado. And the maker’s owner entered its own account on Sunday: Hewlett Packard Enterprise, which owns Juniper, told ABC News that the device worked as the installer had configured it and did not fail or act improperly. That statement, if it stands, points toward configuration; the FAA has not said so, and the desk prints the pointer as the maker’s, not the record’s.

The open questions, as the record leaves them. First, the initiating fault: installer error, misconfiguration, or defective equipment—the Secretary’s own framing on Saturday was “whether we had incompetence—meaning, was it user error or installer error from L3Harris? Or did they install a faulty router?” As of this report, no finding has been published and no date given for one; the House Transportation and Infrastructure Committee’s chairman and ranking member have said jointly that the committee will thoroughly review the issues that led to the stop, committee staff were to be briefed on Monday, and two of the committee’s Democrats expect answers from the Administrator at a hearing in early December or sooner. Second, the containment question: why the flood could not be isolated before it reached 17 of 22 centers, and what the disabled broadcast function is, in technical terms—no official account in the record answers either. Third, the discrepancy in the government’s own numbers: approximately 60 routers against 69, and new routers against routers bought in 2001. Fourth, the question of the hour of the work. The El Dorado replacement was done in the middle of a Friday afternoon, at the edge of the week’s heaviest departure bank. The FAA’s written account answers the question directly: maintenance at major facilities is typically done at night, but replacements at remote facilities “regularly” happen during the day and have done so “for decades, without any impacts”; the Secretary added that the telecom provider required the change because the copper line was being retired; and both the agency and the Secretary state that the timing of such work will now be re-evaluated. Whether re-evaluation becomes a rule moving such work to the night hours is not in the record. Fifth, recurrence: the Secretary declined on Sunday to guarantee that it cannot happen again, and said on Saturday that more outages like this one should be expected until the old network is replaced. The President, on Sunday, blamed the state of the system on the previous administration and urged Congress to approve the money for a new one quickly; New York’s governor, on Saturday, called for the Secretary to resign or be removed, and for Congress to investigate; the House Speaker said no investigation is needed and that further investment is. The desk files the three positions as positions. The record underneath them is the timeline, the mechanism, and the counts, and it is the same record for all three.

5. The SMART Question

The question exists because of a date. On Tuesday, November 10, 2026—three days before the stop—the FAA announced that SMART, its AI-supported planning tool, was expanding to the airspace of all 22 Air Route Traffic Control Centers. SMART stands for Strategic Management of Airspace, Routes and Trajectories. It was unveiled on September 21 around Washington, D.C., and the expansion took it nationwide. When the centers whose airspace SMART had just entered lost their radar and radio stability on Friday afternoon, the question whether the new tool was connected to the failure was asked the same day, in the open, and it has been answered the same way by every party to it. This section files the question and the answers together, each with its attribution, and adds nothing to either.

What SMART is, on the record. The tool is built by Air Space Intelligence, a Boston company, under a contract awarded in June 2026—a 12-year award of roughly $875 million covering SMART and the Command Center’s new data backbone. On the FAA’s own description, SMART is a cloud-based platform based in a single location near Washington, where a team analyzes flight schedules, weather forecasts, and hundreds of other data streams and produces a shared picture of where aircraft are going, how much traffic the system can handle, and where congestion or weather could cause problems, before flights depart. It makes recommendations—routing, departure timing, arrival sequencing—to human beings. The Secretary’s description of the interface is the plainest: SMART “only gives a person a prompt, and the human being will actually decide to accept or deny that prompt.” The FAA cites 180 rerouted flights since September 21, with about ten minutes and about eighty gallons of fuel saved per flight on average. The oversight file on the tool predates the failure: in October, House Democrats wrote to the Government Accountability Office requesting an audit of the new software, raising concerns about safety and guardrails, and the controllers’ union has said it was not consulted in designing the system. Those facts are filed here because they are the tool’s record, not because they bear on Friday; nothing in them speaks to the ground stop, and the desk does not make them speak to it.

The denials, in the order they were entered. First, the company’s, at 5:34 p.m. on Friday, while departures were still stopped: “The current ground stop is not based on or a result of the SMART system, any product of ASI or our integration with FAA and ATC.” Second, the Secretary’s, at the White House on Friday night, in answer to the question directly: the SMART system “isn’t plugged in to STARS and ERAM”—the terminal and en route automation systems—“but that SMART system isn’t plugged in… By the way, if it was, I would tell you that, but this is not even plugged in.” His conclusion: “There was no possible way SMART would have impacted what happened today.” Secretary Mullin, the same night, said there was no indication the communications problems were linked to the new system. Third, the Administrator’s, on Saturday: investigators had seen “zero evidence of anything nefarious, and zero evidence of anything AI-related.” Fourth, the Department of Transportation’s, on Saturday evening: the instability was triggered by routine maintenance on the existing telecommunications network, not by the modernization programme and not by the SMART tool. Fifth, the FAA’s written account, on Sunday, in its own questions and answers, under the question “Was this caused by SMART system?”: “No. The two systems are not related. SMART is a data aggregator, and it does not feed information into the air traffic control system.” The Associated Press, reporting the denials on Monday, added the sentence the desk adopts as this section’s close: the timing has still drawn questions. The timing is a fact; the denials are facts; the record contains no third thing, and the report prints none.

6. The Backdrop: a Network Under Renovation

The network that failed was a building under renovation with the tenants still in it. FTI has been flagged as a weakness for longer than some of its equipment has been old: federal auditors have marked the network as an outage risk for more than fifteen years, and the Washington Post reported on Saturday that the FAA had been warned for years about the risks of the repair work itself. The network has failed nationally before, at the same seam. On November 19, 2009, an outage of the FAA’s telecommunications hit 129 of the nation’s air traffic facilities and grounded hundreds of flights; a Transportation Department inspector general review followed in June 2010. The cause found for 2009, in the record of that event, was a contractor’s misconfigured router at the junction between the old network and the new one. The desk files the precedent without pressing it: whether Friday’s fault sits in the same place is precisely what the present investigation has not determined.

The replacement program, as the auditors filed it. The governing document is the Government Accountability Office’s report of September 14, 2026, GAO-26-107992, on the Brand New Air Traffic Control System—BNATCS—the FAA’s accelerated modernization. Its findings, in its own terms: Phase 1, which modernizes communication, surveillance, weather, and training systems, “is to be fully completed by December 2028.” Congress appropriated $12.5 billion in July 2025, most of it for Phase 1. Phase 2, new automation to track aircraft and optimize traffic, carries an FAA estimate of about $10.2 billion and no start or completion date. As of May 2026, the FAA reported replacing 2,560 of 5,170 “old and frail copper wires” with fiber-optic cable—49.5 percent of the wire count, a figure the desk prints beside the contractor’s “more than 70 percent complete” and the administration’s “68 percent of the legacy copper network replaced,” because the three percentages measure three different things: wires replaced, program completion, and network share. And the planning findings, which are the audit’s center: the FAA “has not developed a comprehensive and well-documented lifecycle cost estimate” for the program, and “has not developed an integrated master schedule for phase 1. Instead, FAA has 11,389 individual project schedules that are not integrated.” The money underneath the program has been flat for a generation: by the FAA’s own account, funding for its Facilities and Equipment account hovered around $3 billion a year across fifteen years, costing it roughly $1 billion in purchasing power. The Administrator told a House appropriations subcommittee in September that Phase 1 is costing $16 billion—against the $12.5 billion appropriated, with the gap, on the trade coverage of the audit, covered from the agency’s own facilities and equipment budget. In August 2026, trade reporting states, the FAA selected AT&T to build the replacement network, due by the end of 2028; L3Harris retains the surveillance work, the 700-plus ADS-B ground stations, under a contract that can run to 2045. The desk files the AT&T selection at the grade of its single trade source.

The money argument, as it stood on Monday. The Secretary asked Congress for $50 billion on Friday night—“I want $50 billion”—and repeated the figure through the weekend as the price of a complete overhaul, with a new network possible by the end of 2028 or early 2029. Against the ask, two House Democrats filed the spending record: of the $12.5 billion Congress approved last year, the Department of Transportation has awarded contracts worth $4.3 billion and spent $1 billion. Both figures are now in the record, and the argument between them—whether the failure argues for more money faster or for accounting for the money already given—is the argument Congress will have, in a hearing expected in early December. The desk’s file closes the backdrop where the auditors left it in September: an ambitious program, a December 2028 date, 11,389 unconnected schedules, and a network whose oldest parts were still carrying the nation’s radar and radio traffic on the Friday afternoon the renovation reached the wrong router.

7. The Recovery: Saturday, Sunday, Monday

Saturday was the digging out. The system itself was declared back at full operations at Saturday’s news conference; the airlines were nowhere near it. Aircraft and crews were out of position across the country, and the advisory record shows the carriers managing the wreckage with the same instrument that had held the nation the day before: through Saturday morning and afternoon, United asked for and received a ground stop on its own flights into O’Hare, Delta into LaGuardia, Southwest into San Diego, Delta again into LaGuardia—each marked as the airline’s request, each holding only that carrier’s flights. The last of Friday’s own measures, the extended retention of flight plans ordered for the “nationwide comms failure,” was withdrawn at 7:42 a.m. Saturday; from that hour, nothing the Command Center had put in place for the outage remained in effect, and the day’s remaining disorder belonged to the airlines’ schedules rather than to the network. The counts are in the third section: the Administrator’s morning forecast of 4,000 to 5,000 cancellations out of about 41,000 scheduled was passed by mid-afternoon, and FlightAware’s day closed at about 6,700. The waivers tell the same story in the airlines’ own expectations: as of Sunday, American’s travel alert ran through Wednesday, November 18; Delta’s and Southwest’s through Monday; United’s and JetBlue’s through Sunday.

Sunday was the slope down. Traffic in the arrival system on Sunday evening ran about 30 percent above Saturday’s at the same hour and about level with Thursday evening’s, before the outage—the system’s own measure of itself returning. Cancellations fell to roughly 1,900 to 2,400 on FlightAware’s snapshots, or just over 1,300 on Cirium’s count, a 5.2 percent rate against the 3.4 percent baseline. The Secretary, on the morning programs, set the expectation he had set on Friday night—normal by the end of Sunday or on Monday—and declined to guarantee it. Delta said most of its flights were operating normally. No large carrier had said by Sunday night that it was fully back to normal, and American’s alert running through Wednesday said the quiet part for the industry.

Monday was, by the counts, a normal day with an asterisk. By about 6 p.m., FlightAware counted 344 to 349 cancellations across the United States—against about 2,300 on Sunday, about 6,700 on Saturday, and the Friday figures in the ladder—and some 4,800 delays. The asterisk is the next section. The recovery’s human arithmetic stands in the record at Cirium’s estimate of 1.5 million people stranded by the cancellations, and its financial arithmetic at the Secretary’s Monday estimate, given on CNBC, of an impact in the billions of dollars. The week it landed in is the week before Thanksgiving travel; the Secretary said Friday night that he expected the system fully back before the rush, and unnamed officials told the Associated Press the same. On the counts, by Monday evening, it was—the network’s failure had closed. What remained open was everything in the fourth section: the fault, the containment, the numbers that do not reconcile, and the hour of the work.

8. Monday’s Separate Failure: United Airlines

This section is a different event, and the report keeps it in a section of its own because the record does. On Monday, November 16, in the middle of the afternoon, United Airlines stopped its own departures. The failure was inside the airline: a companywide, web-based outage that hit several pilot applications, including crew communications, and cut off the airline’s ability to provide final aircraft weights for departing flights. The source for the mechanism is a memo of the Air Line Pilots Association, the union representing United’s pilots, seen by Reuters and reported by ABC News and others: without final weights, a flight cannot complete its weight-and-balance calculation, and without that calculation it cannot leave the gate—the airplanes, the crews, and the airspace above them all serviceable, and the number that releases them missing. Pilots first reported the trouble to controllers at about 3:30 p.m. Central time. United’s statement, when it came, was short: “We have resolved a technology issue that delayed flight departures,” with “no indication that this was a cybersecurity-related issue.” The resolution came at about 6 p.m. Eastern. The duration, as printed, runs from about one hour in the airline’s framing to about ninety minutes to about one hour and forty minutes measured from the first pilot reports to the resolution; the measures differ because they start their clocks at different points, and the desk prints the range rather than choosing. Flights already in the air were not affected, and neither was United Express, the regional network.

Monday’s counts, with the same discipline as Friday’s. FlightAware’s snapshots: 418 United flights delayed and 21 cancelled at 5:30 p.m.; by 6:50 p.m., 645 delayed—about 20 percent of United’s schedule—and 25 cancelled. How many of those delays belong to Monday’s outage and how many to the Friday recovery still running underneath them is not separable in the counts, and the outlets that printed the figures said so. One figure from Monday night requires its own correction, because it travelled: NBC’s evening segment on the outage displayed 4,400 flights delayed and 342 cancelled. The 342 was not a United figure. It was the all-carrier United States cancellation count for Monday at that hour—FlightAware’s 344 to 349 by about 6 p.m.—printed beside the United story, where it read as the outage’s toll. United’s own Monday cancellations were 25. The desk files the correction in the open, in the place where the figure is most likely to be met.

The separation, on the record. The FAA stated on Monday that the problem was a United technology issue, “not an FAA equipment issue,” and that it had no indication Monday’s failure was connected to Friday’s shutdown; the Secretary called it “not an FAA system-wide problem.” United has not named the system that failed, and has not said what caused the outage. One precedent stands in the airline’s own record: in August 2025, United briefly grounded its flights over a problem with Unimatic, the system that supplies the flight information used in weight-and-balance calculations. Whether Unimatic was involved on Monday is not established, and the desk does not establish it. The Federal Bureau of Investigation’s place in this week’s record belongs to Friday—the examination of the El Dorado router—and nothing in the record puts the Bureau inside United’s failure; a Monday news segment’s statement that “the FAA and FBI are investigating,” carried over pictures of the United outage, is filed here as the segment’s framing, against the agencies’ own placement of the Bureau’s work. Two failures, four days apart, in two different systems owned by two different parties, with no indicated connection between them: that is the whole of the relation, and the report states it once, here, and does not state it again.

9. What the Record Does Not Settle

The desk closes the file with its cocked hat printed, as the paper’s method requires. The record establishes the afternoon: the advisory times, the stop’s length, the flood’s rate, the two sites, the blocking that ended it, the safety of every aircraft airborne, and the counts as each counter made them. It does not establish the initiating fault—installer error, misconfiguration, or a defective device—and the parties’ statements point in different directions: the Secretary holds the fork open, and the device maker’s owner says the device did as it was configured. It does not establish whether Amarillo failed alone or was set off. It does not explain why one site’s flood reached seventeen centers, nor name the hour the flood was blocked, nor reconcile the government’s two router counts—about sixty, and sixty-nine—nor its two descriptions of the same machines, new and bought in 2001. It does not give a single agreed cancellation total for Friday, and this report has not manufactured one. It does not say whether remote-site replacement work will move to the night hours, only that the timing will be re-evaluated. It does not say when the investigation will report; a congressional hearing is expected in early December. And it does not connect Monday’s failure inside one airline to Friday’s failure inside the nation’s network, because the agencies that own both records state that no connection is indicated. The file stands open at those points, and the paper leaves it open, as a log leaves a question it cannot answer: printed, dated, and unsettled.

The Filed Clippings

Everything the desk read for the report above, clipped and filed as a bureau files it. Each item is stamped subjective, a source with a position, printed because it is useful, and described by what it says. The desk’s report is the only part of this file stamped objective. Fourteen items.

Subjective
Federal Aviation Administration—“Nationwide Ground Stop: Questions and Answers” (November 15, 2026)

The agency’s settled written account, and this report’s primary document: the flood at 1,000 connection attempts every four to five seconds; the two sites, Amarillo and El Dorado; the blocking that stabilized the network; 17 of 22 centers; approximately 3,500 aircraft landed safely; no indication of a cyberattack; SMART described as a data aggregator that does not feed the air traffic control system; approximately 12,462 cancellations on Friday, 9,082 by major airlines, with the caution that not all were directly caused by the stop; the daytime-work answer and the promise to re-evaluate it; and the broadcast function disabled on approximately 60 similar routers.

Subjective
IFR Meter—“Nationwide ground stop of 13 November 2026: cancelled after four hours and 22 minutes” (running log, November 13–16, 2026)

The advisory record kept as it arrived: Advisory 111’s text and 3:54 p.m. order, the Chicago stops that preceded it, the diversion machinery, Advisory 120’s recovery path, Advisory 005’s 8:16 p.m. cancellation, the airline-requested stops of the night and of Saturday, and a running collation of the officials’ statements and the counters’ figures across all four days, with the log’s own discrepancies flagged in its text.

Subjective
Government Accountability Office—GAO-26-107992, “Air Traffic Control Systems: Ambitious New Modernization Effort Needs to Improve Cost and Schedule Planning” (September 14, 2026)

The auditors’ report on the Brand New Air Traffic Control System: Phase 1 to be fully completed by December 2028; the $12.5 billion appropriated in July 2025; Phase 2 estimated at about $10.2 billion with no dates; 2,560 of 5,170 copper wires replaced as of May 2026; no comprehensive lifecycle cost estimate; and, in place of an integrated master schedule, 11,389 individual project schedules that are not integrated.

Subjective
TechCrunch—“Transportation secretary blames ‘routine’ equipment upgrade for grounding flights nationwide” (November 14, 2026)

The Saturday news conference as the trade press reported it: routine maintenance, copper upgraded to fiber by L3Harris at a remote location, a router overloading a critical communication system; Cirium’s more than 8,200 Friday cancellations and at least 3,000 more on Saturday; the third nationwide ground stop of the century; and the controllers’ union’s statement that it was not consulted in designing the new planning tool. The acting chief technology officer’s account of finding the two sites—“pinging our network at a very high rate”—is carried in this coverage.

Subjective
“Anatomy of the FAA Ground Stop: One Backbone, One Bad Change” (November 15, 2026)

A technical compilation, with confirmed facts attributed and analysis labeled: the stop’s 3:54 p.m. to 8:15 p.m. span; instability beginning about 3:30 to 3:36 p.m., with the Secretary’s 3:36 onset set against another outlet’s use of the same figure for the stop itself; FTI’s TDM core and the 2002 Harris contract; the third nationwide stop in the country’s history; and the observation that within about twenty hours the official explanation shifted from overloaded legacy routers to a contractor’s cutover to an admission that the fault itself has not been determined.

Subjective
“FAA Ground Stop: Years of Network Warnings” (November 15, 2026)

The warning record assembled in one place: auditors flagging the FTI system as an outage risk for more than fifteen years; the Administrator’s Saturday statement that every agency in the government for better than two decades has flagged FTI as a critical weakness; the Washington Post’s reporting that the FAA had been warned for years about the risks of the repair work; the September GAO report’s place in the timeline; and the plan to replace 5,170 copper lines and move nearly 30,000 services to Internet Protocol by September 2030.

Subjective
“FAA Says L3Harris Owns the Network Involved in the Nationwide Ground Stop” (November 13–15, 2026)

The accountability frame, filed as a source with a position: the Secretary’s statements that L3Harris has served the FAA for more than twenty years and owns the network used to control the airspace; the 1,000-plus connection attempts every four to five seconds; Reuters’ 17 of 22 centers and its more than 20,000 flights disrupted; the Department’s 3,500 aircraft airborne and more than 8,000 flight operations resumed; L3Harris’s statement that the conversion is more than 70 percent complete, which does not identify the initiating fault; and this file’s reading of Advisory 005 as cancelling the stop at 8:19 p.m., against the advisory record’s 8:16.

Subjective
“What to know about the FAA’s SMART AI system” (November 16, 2026)

The SMART file in one article: the $875 million system expanded nationwide on November 10, three days before the stop; the company’s and the officials’ denials, including the Secretary’s and the Administrator’s restatements on Saturday; the system’s single location near Washington and its advisory character—recommendations to humans, not control of aircraft; and the October letter in which House Democrats asked the GAO to audit the new software, raising safety and guardrail concerns.

Subjective
The Times—“FBI examines router at centre of US flight shutdown chaos” (November 17, 2026)

The Bureau’s place in the record: the FBI assisting aviation experts in examining the El Dorado router, which incorrectly sent hundreds of connection requests a second to a broad network of other routers; 18,500 flights cancelled and 30,000 delayed in the paper’s summation; the Secretary’s Saturday statement that the installation was made by an Air Force veteran working for L3Harris; the routers under examination at a laboratory in Florida; and, in its second half, Monday’s United failure, with the air traffic control audio of aircraft unable to receive their take-off numbers.

Subjective
“FAA Names Two Potential Sites in Ground-Stop Probe” (November 15, 2026)

The Sunday account in compact form: the FAA naming Amarillo and El Dorado; the data flood against legacy equipment; the 12,462 Friday cancellations with the agency’s caution; the Associated Press’s Cirium figures for Sunday—just over 1,300 cancellations among more than 26,000 scheduled, a 5.2 percent rate against the 3.4 percent pre-Friday baseline; and Reuters’ FlightAware figures for the same Sunday, more than 2,300 cancellations, printed beside Cirium’s with the note that the figures cover different measures from separate providers.

Subjective
“L3Harris Built the Wires Under U.S. Air Traffic Control. Now It’s Handing Them to AT&T” (November 2026)

The network’s corporate history and its handover: L3Harris operating FTI since 2002 across more than 4,400 sites; departures from 3 p.m. to 9 p.m. on November 13 running 72 percent below the same hours the day before, about 11,000 flights that never left; the FAA’s August 2026 selection of AT&T to build the replacement network, due by the end of 2028; L3Harris retaining the surveillance contract for the 700-plus ADS-B ground stations, which can run to 2045; and the trade speculation, labeled as such in the article, about a network change that failed to hand traffic to its backup.

Subjective
“US air travel recovery stumbles on United pause as funding fight brews” (November 17, 2026)

Monday’s wrap, wire-sourced: air travel largely back to normal after more than 18,000 cancellations over three days; the Secretary’s CNBC estimate of an economic impact in the billions and nearly 28,000 flights delayed across the disruption; the United outage resolved by 6 p.m., not affecting other airlines; United’s 645 delayed and 25 cancelled at 6:50 p.m. by FlightAware; the all-carrier Monday figures of 344 cancellations and 4,800 delays at 6 p.m.; and the statement, carried in the article, that it was unclear how many of Monday’s disruptions the outage itself caused.

Subjective
NBC News—“Some airports beginning to run out of space for planes amid nationwide ground stop” (November 13, 2026, as carried in summary)

The parking fact of Friday evening: a source with direct knowledge telling NBC News that Reagan National Airport had reached full capacity and could no longer accommodate additional planes, that Chicago O’Hare was nearing capacity, and that the shortage of parking could force aircraft to divert to airports where space remained. Filed from the summary of the report; the full text stands with its publisher.

Subjective
“United Airlines Fixes Tech Outage That Delayed Flights Nationwide” (November 16, 2026)

Monday’s event as the wire carried it: United’s resolved technology issue and its no-cyberattack statement; the pilots’ union’s description of a systemwide operational outage affecting pilot applications, crew communications, and the provision of final aircraft weight information; the problem fixed about an hour and a half after pilots first reported trouble; FlightAware’s 418 United delays late Monday afternoon; and, in the same article, the all-carrier figures—more than 340 flights still cancelled across the United States late Monday afternoon, and 347 by 6 p.m., down from about 2,300 on Sunday, 6,700 on Saturday, and more than 9,000 on Friday.