Microsoft attributed the November 12, 2026, Microsoft 365 outage to a latent software defect in its networking gateway, a shift in Teams traffic and reduced gateway capacity. In an account published by Microsoft executive Charles Lamanna, the company said those conditions combined to trigger the failure; the defect involved lock contention inside an internal process. Lamanna’s account traced how the gateway problem spread and complicated recovery.
What caused the Microsoft 365 outage?
Microsoft identified three conditions: a latent gateway-software defect, more traffic moving onto the primary gateway after Teams failed over, and nearly 14% of gateway capacity already unavailable because of a separate issue the previous day. Microsoft said that capacity loss alone had not affected service because the remaining infrastructure had enough headroom.
Lock contention occurs when operations compete for access to a shared resource and must wait for one another. Microsoft said the defect made gateway servers unavailable under certain load conditions. The change associated with it began deployment in July 2026 and had been deployed globally for more than three months without observed impact before the outage.
The defect was specific to Microsoft 365 networking gateway servers, according to Microsoft; the company said it did not apply to general Windows deployments.
Why did testing miss the defect?
Microsoft said its test environments used different traffic patterns and a different mix of server types, or SKUs, than production. Tests tended to focus on weaker machines with fewer processor cores, while the contention was worse on newer machines with more cores. That mismatch meant the test workload did not reproduce the conditions in which the defect surfaced.
How did the gateway failure cascade?
About 15 hours before the outage, Teams failed over from an alternative gateway to the primary gateway, which also supports Exchange Online. Microsoft described the failover as a normal resilience operation. The resulting shift put more load on the primary gateway while nearly 14% of its capacity was unavailable for the separate prior-day issue.
When servers became unresponsive, automatic routing sent users to other healthy endpoints. That spread the same traffic across fewer servers, making the defect more likely to recur on the remaining endpoints. A separate problem also prevented the gateway’s usual partition boundaries—designed to contain failures—from stopping the cascade across regions. Microsoft said the defect affected the server fleet in less than two hours; at the low point, more than 80% of servers were unresponsive.
Failover can move traffic away from a failing endpoint, but it cannot protect a service if the remaining endpoints encounter the same defect under the shifted load.
Why did Exchange Online have a different recovery path?
Microsoft said Teams, OneDrive, SharePoint and Office apps could fail out of the affected gateway, while Exchange Online could not. That difference constrained how traffic could be redirected as engineers worked to restore service.
Servers returning to service were overwhelmed by queued traffic, and the usual automated recovery and rollback mechanisms were not designed for such a broad unresponsive state. Microsoft said engineers returned the code path to a known-good state, then restored service in batches by taking groups of servers offline, applying the update and restarting them.