Outlook Downtime: Latent Bug in Gateway Servers
Microsoft’s email service was down for nearly 24 hours after a dormant defect in Windows servers was triggered by a routine failover and reduced capacity.
By Chinedu Schneider ·

Microsoft’s Outlook email service went down for nearly 24 hours last week, marking one of the company’s most severe outages.
The failure started in Switzerland when Exchange Online servers began going offline. The disruption spread to Mexico and then globally as failover systems shifted traffic. Within two hours, more than 80 percent of Exchange servers were unresponsive. Microsoft labeled the incident as a sev0, its highest severity level, and mobilized hundreds of engineers.
Engineers initially investigated potential cyberattacks and AI-generated code errors. Crash reports and kernel dumps eventually ruled out an external attack. The team found a latent bug in Microsoft 365 networking gateway servers that had existed since July. Charles Lamanna, head of Microsoft’s Copilot, Agents, and Platform team, stated that a latent defect caused servers to become overwhelmed under specific load conditions.
Three factors triggered the dormant defect. Fifteen hours before the outage, Microsoft Teams failed over to the main gateway supporting Exchange Online. This increased the load and changed traffic patterns. At the same time, roughly 14 percent of the gateway’s server capacity was out of service due to an unrelated issue from the previous day. The combination of the bug, the failover, and the reduced capacity caused the collapse.
Recovery was complicated by the nature of the failure. As servers returned to service, they were immediately overwhelmed by waiting traffic and became unresponsive again. Automated recovery processes were not designed for such a broadly unresponsive state. A running bridge call for engineers hit its maximum of 1,000 participants. Overlapping interventions from multiple teams sometimes complicated coordination. It took nearly 24 hours to stabilize the Exchange servers.
The preliminary post-incident report describes a flaw affecting select Windows servers due to lock contention inside an internal Windows process. The bug did not affect regular versions of Windows because it stemmed from a fix built specifically for the gateway servers. Lamanna acknowledged testing gaps, noting that the contention issue is worsened on machines with more cores and newer hardware, while load testing tended to focus on weaker hardware with fewer cores.
Microsoft plans to publish a final post-incident review later this week. The company will need to invest more heavily in its testing processes to prevent a similar sequence of events from knocking out Outlook for millions of people again.
More Tech · National news · Report a business · Got a Story?






