· via The Verge
Latent Microsoft 365 gateway bug, Teams failover and reduced capacity caused day-long Outlook outage
A dormant networking bug present since July, a Microsoft Teams failover and roughly 14% lost capacity knocked out more than 80% of Exchange servers, leaving Outlook users offline for nearly a day.

How the outage unfolded
According to The Verge's inside account, the incident began barely an hour into a seemingly normal working day last Thursday, when Exchange Online servers that power Outlook started dropping offline in Switzerland. Microsoft's failover systems tried to shift email traffic elsewhere, but that only spread the fault — first to Mexico, then to countries around the world. Within roughly two hours, more than 80 percent of Exchange servers were unresponsive, and employees, businesses and many governments were left without Outlook email for close to 24 hours.
The Verge reports that Microsoft logged the event as "sev0" in its internal incident tracker, the highest severity rating, and that hundreds of engineers mobilised to diagnose it. The failover machinery that normally protects the service ended up amplifying the outage by pushing traffic onto servers that then failed too.
A dormant bug meets a traffic spike
In the early hours, engineers pored over crash reports, trace logs and kernel dumps. Some servers had slipped into maintenance mode and stopped responding under unusually high CPU load, and every attempt to bring them back was immediately swamped by queued traffic. Because Exchange servers boot their operating system over the network via Preboot Execution Environment, the wave of reboots overwhelmed the systems monitoring them, creating a loop in which servers repeatedly tried and failed to return.
Microsoft also treated a possible attack as a live theory, liaising with DDoS specialists while observers outside speculated about a cyberattack or careless AI-generated code. Crash reports and logs ruled an attack out, The Verge says.
The real cause was a latent defect in the networking stack of Microsoft 365 gateway servers, present since July without any visible impact. Microsoft's preliminary post-incident report attributes it to "lock contention inside an internal Windows process." Because the flaw came from a fix written specifically for these gateway servers, regular versions of Windows are unaffected.
Charles Lamanna, who leads Microsoft's Copilot, Agents and Platform organisation and began overseeing engineering for Office, Outlook and Teams last month after Ryan Roslansky left the company, wrote on X that "a latent defect caused servers to become overwhelmed and unresponsive under certain load conditions."
Two other incidents made the defect visible. Fifteen hours before the Outlook outage began, Microsoft Teams failed over from an alternative networking gateway onto the main gateway that supports Exchange Online, raising its load and changing its traffic patterns. At the same time, around 14 percent of that gateway's server capacity was out of service because of an unrelated issue from the previous day. That combination triggered a bug that had been sitting dormant for months.
Microsoft also admits the bug escaped its testing for a mundane reason: the contention worsens on machines with more cores and newer hardware, while its load testing leaned on weaker machines with fewer cores.
Why recovery took nearly 24 hours
Lamanna acknowledged that restoring service was unusually hard. "Recovery took longer than usual because as servers returned to service, they were immediately overwhelmed by waiting traffic and became unresponsive again," he wrote. Automated recovery processes were not built to restore service from such a broad unresponsive state, and because the defect had shipped months earlier, normal rollback, flighting and feature-flag tools could not quickly undo it.
The Verge's sources add that overlapping interventions from multiple teams sometimes complicated coordination, and that the continuous bridge call hit its 1,000-participant maximum at one point — a measure of how many teams were drawn into the response, which began at peak traffic in the US. Microsoft says a final post-incident review will be published later this week.
Why it matters
The outage is a rare, detailed case study of cascading failure in a hyperscale cloud, and three lessons stand out. Redundancy can backfire: the failover systems did exactly what they were designed to do and still spread the failure. Testing can miss production reality: a bug invisible on weaker hardware proved catastrophic on newer, many-core servers under real traffic. And recovery automation has limits: when an entire fleet goes dark because of a change made months earlier, standard rollback tooling offers little help. For the businesses and governments that run on Exchange Online, the episode is a reminder that continuity planning for cloud services still matters — and for Microsoft, it puts fresh pressure on how it tests and ships changes to the infrastructure behind Microsoft 365.
- #microsoft-365
- #outlook
- #exchange-online
- #cloud-outage
- #post-incident-review