Microsoft's Exchange Online servers, which power Outlook email, went offline for nearly 24 hours, according to a report from The Verge. The outage began in Switzerland, spread to Mexico and then to other countries, and left employees, businesses and governments without Outlook email for much of that time.
A source told The Verge that Microsoft classified the incident as "sev0" in its internal tracking system, the highest severity level. Within a couple of hours, more than 80 percent of Exchange servers were unresponsive. The report says Microsoft's failover systems, designed to shift traffic to healthy servers, spread the problem.
Charles Lamanna, leader of Microsoft's Copilot, Agents, and Platform team, wrote on X: "A latent defect caused servers to become overwhelmed and unresponsive under certain load conditions." Microsoft's preliminary post-incident report attributes the bug to "lock contention inside an internal Windows process" affecting select Windows servers. The report says the bug stemmed from a fix prepared for Microsoft 365 networking gateway servers and does not affect regular Windows versions.
According to Lamanna, the bug had been present on those gateway servers since July without causing impact. Microsoft said its testing did not catch it because the labs focused on weaker hardware with fewer cores, while the contention issue worsens on machines with more cores and newer hardware.
Lamanna wrote that two earlier events triggered the bug. About 15 hours before the outage, Microsoft Teams failed over to the main gateway supporting Exchange Online, increasing its load and changing traffic patterns. At the same time, about 14 percent of that gateway's server capacity was out of service because of an unrelated issue from the previous day.
Lamanna also said recovery took longer than usual because servers that returned to service were immediately overwhelmed by waiting traffic and became unresponsive again. He wrote that Microsoft's automated recovery processes were not designed for restoring service from such a broad unresponsive state. The Verge reported that an internal bridge call reached its 1,000-participant maximum during the response.