At around 3:30pm Eastern time on Friday November 13, the Federal Aviation Administration (FAA) ordered every departing flight in the United States to stay on the ground. Nearly 3,500 aircraft already in the air landed safely, and the busiest airspace in the world stood still for more than four hours. By the time the ground stop was lifted at 8:15pm, more than 9,000 flights had been cancelled.The disruption ran on through the weekend and eventually affected upwards of 20,000 flights. It was the longest nationwide ground stop in the United States since 11 September 2001, and investigators believe it began with a single router.L3Harris, the defence contractor that runs the FAA's telecommunications network, had been installing new equipment at an FAA site in Arkansas. The work was part of a programme to replace legacy copper lines with fibre optic cable. The new router began flooding the network with more than 1,000 connection attempts every four to five seconds.Within minutes the instability had spread to 17 of the FAA's 22 Air Route Traffic Control Centres, the facilities that manage aircraft at cruise altitude, and on into several of the approach facilities beneath them. No centre went completely dark and some radio frequencies and radar feeds stayed up. Controllers were still left working with intermittent gaps in communications and radar.Engineers spent around 45 minutes watching the system behave erratically without finding the source. With no root cause and no assurance that every centre would hold, halting departures nationwide was the correct decision, and the safeguards worked exactly as designed. A second router in Amarillo, Texas, more than 500 miles from the Arkansas site, also began causing problems, and why both failed is now the central question for investigators.L3Harris eventually blocked traffic from both locations and the network stabilised. A cyberattack has been ruled out. Transportation Secretary Sean Duffy says investigators are still establishing whether the fault lies with the equipment or its installation.The most telling aspect of the episode is that the work to modernise the system triggered the outage.In September, the Government Accountability Office published a review of the FAA's modernisation programme that documented the replacement of copper wiring dating back to the 1960s. As of May, 2,560 of 5,170 of those connections had been switched to fibre, and L3Harris says it has completed more than 70% of its planned conversion. Anyone who has worked on critical infrastructure knows the migration phase carries the greatest risk. Air traffic control has no maintenance window and cannot be switched off for a weekend while new equipment goes in. Every upgrade therefore takes place on a live network that handles more than 50,000 flights a day.The age of that network is well documented. In 2025, the FAA's own leadership told Congress that parts of the system still ran on floppy disks, paper flight strips and computers using Windows 95, and pledged to retire all three. A 2023 FAA assessment of 138 air traffic control systems rated 51 as unsustainable and a further 54 as potentially unsustainable. Four of the most critical systems were more than 30 years old, and none had a scheduled replacement before 2030.Friday's outage is the third serious failure in under four years. In January 2023, a contractor unintentionally deleted files in the NOTAM system and triggered the first nationwide ground stop since 9/11. In September this year, primary and backup data links failed at the Philadelphia facility as world leaders arrived for the UN General Assembly. Flights into New York, Philadelphia, Boston and Washington were halted for eight hours, and around 9,500 flights were delayed or cancelled.The operational fallout from a four-hour stoppage lasts for days. Aircraft end up in the wrong cities, crews exceed their legal duty limits and hub connection banks fall out of sequence, which is why airlines were still issuing travel waivers well into the following week. More than 1.5mn passengers were affected, and the impact reached far beyond the United States. Departures from Europe to American airports were suspended for a seven-hour window. Long-haul operations suffered most, because every cancelled transatlantic departure also takes out the return sector and often the aircraft's next rotation.Three lessons stand out. The first is network architecture. A single router at a regional site should never be able to destabilise 17 en route centres, and a network carrying safety-critical traffic needs hard boundaries between its parts so that a fault in Arkansas stays in Arkansas. The second is change control. Work on live air traffic infrastructure demands thorough validation before deployment and a staged rollout after it.FAA staff are now examining both routers with L3Harris at its Florida laboratory, and the industry will want to know what testing took place before installation. The third is delivery. Congress approved $12.5bn for modernisation last year, the White House is seeking a further $30.1bn and Duffy is now calling for around $50bn in total. The GAO's September report found that the FAA had yet to produce a comprehensive lifecycle cost estimate for its new system. Funding on this scale requires a credible plan, firm milestones and contractors held to account. Duffy has promised that accountability, and airlines will be watching for it.The author is an aviation analyst. X handle: @AlexInAir.