JupiteX Get the app
Science & Technology17 Nov 2026 · about 7 min

Cascading Failure Deep Dive: How a Months-Old Bug and One Router Took Down Email and Air Travel

The brief

The Microsoft outage reached beyond email: Exchange Online and Outlook were hit hardest, and Teams, Copilot, SharePoint, and other services were affected. The disruption shows how a fault in shared infrastructure can reach tools people use for different kinds of work. In the United States, the FAA stopped every departing flight nationwide for more than four hours. Air traffic controllers had seen radar and radio inconsistencies, and instability affected 17 of 22 high-altitude traffic centers, as well as several approach control facilities. The stop was lifted at about 8:15 p.m. The two incidents occurred on consecutive days, but officials said there was no sign they were related, and neither had been tied to a cyberattack. The explanations remain preliminary: Microsoft’s final post-incident report was due November 20, and the FAA’s root-cause review was unfinished. The provided article does not detail which individual flights or airports were affected.

01

Which Microsoft services and parts of U.S. air travel were disrupted?

The Microsoft outage reached beyond email: Exchange Online and Outlook were hit hardest, and Teams, Copilot, SharePoint, and other services were affected. The disruption shows how a fault in shared infrastructure can reach tools people use for different kinds of work.

In the United States, the FAA stopped every departing flight nationwide for more than four hours. Air traffic controllers had seen radar and radio inconsistencies, and instability affected 17 of 22 high-altitude traffic centers, as well as several approach control facilities. The stop was lifted at about 8:15 p.m.

The two incidents occurred on consecutive days, but officials said there was no sign they were related, and neither had been tied to a cyberattack. The explanations remain preliminary: Microsoft’s final post-incident report was due November 20, and the FAA’s root-cause review was unfinished. The provided article does not detail which individual flights or airports were affected.

02

How widespread and long-lasting were the outages?

The outages differed in duration and scale, but both disrupted systems used across wide areas. Microsoft 365 email and collaboration services failed for about a day across dozens of countries. The FAA’s nationwide ground stop began at 3:54 p.m. Eastern and was lifted at about 8:15 p.m., keeping every departing U.S. flight on the ground for more than four hours.

At Microsoft, the failure spread through its networking gateway servers. More than 80% became unresponsive after overloaded gateways pushed extra traffic onto others. At the FAA, instability reached 17 of the 22 Air Route Traffic Control Centers that manage high-altitude traffic, along with several approach control facilities.

These figures come from preliminary public explanations, not completed investigations. Microsoft’s final incident report was due November 20, and the FAA had not finished its root-cause review. The article gives the FAA stop’s start and end times, but does not state a precise duration in minutes. It also notes the incidents were not known to be related.

03

What local faults are known to have started each incident, and what is still unconfirmed?

Both incidents had a local starting point, but the explanations are not equally complete. Microsoft’s preliminary report traces its outage to a defect in a networking gateway fix rolled out in July. Under heavy load, the defect caused lock contention that consumed CPU and memory without useful work.

For the FAA, L3Harris was replacing copper circuits with fiber at a site in El Dorado, Arkansas, and a new router was installed. The router began generating about 1,000 connection attempts every four to five seconds across the FAA telecommunications network. The article says the router functioned as configured by the installer, according to HPE, which owns Juniper.

The precise cause of the FAA router’s behavior remains undetermined: the article says it could have been a configuration error or a faulty unit. Microsoft also had unresolved questions: why regional safeguards failed to contain the problem. Both investigations were still preliminary, so these explanations may be refined.

04

What is a cascading failure?

A cascading failure happens when a limited fault spreads through a system instead of staying contained. The key is that the system’s response to the first failure can create extra pressure elsewhere. That matters because a system designed to keep operating can accidentally turn a local problem into a wider outage.

Microsoft’s gateways show the pattern. Some servers stopped responding after a defect caused runaway contention under heavy load. Automatic rerouting shifted their traffic to gateways that were still healthy. That added load pushed those servers into the same failure state, and the outage spread until more than 80% of gateway servers were unresponsive.

The article describes the FAA incident as sharing this shape: a local fault became a system-wide outage because of how the system reacted. Its supplied account says the new router generated connection attempts across the shared telecommunications network, with instability reaching multiple traffic control centers. The FAA’s precise failure stages and root cause were still under review, so the article does not establish the full propagation mechanism.

05

How did automatic rerouting turn failing Microsoft gateways into a wider outage?

Automatic rerouting is meant to keep service going when a gateway stops responding: its work is sent to other gateways. It matters because this protection can fail when the original fault is triggered by heavy load. In that case, shifting traffic may move the pressure rather than remove it.

That is what Microsoft’s preliminary account describes. A defect in a July fix caused lock contention under heavy load, leaving some gateways unresponsive. The system redirected those gateways’ traffic to healthy ones. The added work drove those gateways into runaway contention too, spreading the failure until more than 80% of gateway servers were unresponsive.

The outage was not resolved simply by rerouting or restarting servers. Microsoft recovered by rolling out an update that reverted the defective change. The incident illustrates why traffic movement needs to be considered alongside capacity and the fault’s trigger. The article does not specify what changes Microsoft will make to rerouting after the incident; its final report was still due November 20.

06

Why can restarting overloaded systems cause a thundering herd?

A thundering herd is a sudden rush of requests from many clients at once. It matters during recovery because restarting a server does not guarantee it can immediately handle the work waiting for it. If many clients try again together, the recovering system can be overloaded before it stabilizes.

That happened during Microsoft’s outage, according to the preliminary report as summarized by GeekWire. When engineers restarted servers, clients that had been failing all reconnected at once. The restarted servers were overwhelmed before they could stabilize. The article identifies this pattern as a thundering herd, adding another obstacle to recovery after the gateways had already been heavily disrupted.

The article does not describe a specific change Microsoft plans to prevent repeat surges, and the final report was still due November 20. The episode shows that recovery needs attention as well as failure prevention: restarting machines can trigger another burst of demand. The source does not give details about how clients reconnected or how long the restart-related overload lasted.

07

Why can one misbehaving device disrupt many services when networks share infrastructure?

When many services or facilities depend on the same network, a fault at one point can affect more than the site where it began. The shared network becomes a common path for traffic, so unusual activity from one device can reach or burden parts of the wider system. That is why shared infrastructure can make a local problem consequential.

In the FAA incident, L3Harris was replacing copper circuits with fiber at an Arkansas site and installed a new router. The FAA said it generated about 1,000 connection attempts every four to five seconds across the FAA telecommunications network. Around the same time, controllers saw radar and radio inconsistencies, and instability reached 17 of 22 high-altitude traffic centers and several approach facilities.

The article establishes the repeated attempts and the broad disruption, but does not explain the exact technical path from the router to each facility. It also says the cause—configuration error or a faulty unit—was undetermined. The FAA’s root-cause review was unfinished, so the precise mechanism remains uncertain.

This brief was written by AI from the original reporting and checked by other models. Names, figures and quotes come from the source; read it for full context.

Read more in the JupiteX app

Pulse is free. New stories every 4 hours, each one broken into the questions that explain it.

Or read more news on the web