Monday, 16 November 2026

Microsoft traces its 24-hour 365 outage to a Windows gateway bug

Microsoft's preliminary report says a fix on Microsoft 365 gateway servers locked up under shifted Teams traffic, taking Exchange Online down for 24 hours.

Key points

  • Microsoft says a gateway-only Windows fix caused the 12-13 November outage.
  • Under 2 hours was enough for the fault to spread across regions.
  • The full repair details are due on 20 November.

Microsoft has put a cause on the Microsoft 365 outage that cut most people off from Exchange Online mail between 12 and 13 November: a bug in a Windows networking change that runs only on its own gateway servers. A preliminary incident report went out on 15 November, and Microsoft says the detailed repair plan follows on 20 November.

Mail access was down from 4:20 PM UTC on 12 November until 4:10 PM UTC the next day. Microsoft’s position is that ordinary Windows PCs never carried the bug, so nobody needs to touch a laptop or a tenant setting.

The change had been live on the gateways since July with no sign of trouble. What broke it was a coincidence of two things: Teams traffic had been moved onto the main gateway Exchange Online also uses, and about 14% of gateway capacity was out for a separate reason. With both in place, tasks inside one process began queueing for shared resources, processors and memory filled, and the gateways stopped answering new connections. Microsoft says its pre-release testing ran on slower machines, where the problem barely shows.

The automatic safety response then spread it. Traffic from failing servers was redirected to healthy ones, which began failing too, and within two hours the trouble had crossed regions, leaving over 80% of servers unresponsive at its worst. Servers that were restarted fell over again once the backlog hit them. Teams, SharePoint and OneDrive could route around the damage; Exchange Online, at its size, could not.

The fix was to put the connection handling back as it was before the change. Engineers took big batches of servers out of rotation, patched and restarted them, and brought traffic back gradually, while Microsoft added capacity and stretched the delivery window so waiting mail would not bounce. Admins can look for incidents EX1550543 and MO1550821 in service health, and Microsoft’s support thread points to the first. Mail was still returning slowly when the outage passed 17 hours.

KA

Founder and editor of Tech Prolonged since 2010. A software engineer by profession, Khurram covers Google, Android and Microsoft, and edits the newsroom's coverage. Find him on X at @khurramar.