Microsoft has revealed the culprit of the global failure: a hidden Windows error overloaded the servers

Microsoft has revealed the culprit of the global failure: a hidden Windows error overloaded the servers

Microsoft reported that the global outage of Microsoft 365 services on November 12-13, 2026 was caused by a hidden bug in Windows, due to which the servers responsible for cloud connections were overloaded. The preliminary incident report explains that the software change caused traffic redistribution and reduced server capacity, causing widespread outages. The Exchange Online service was most seriously affected.

Users in more than 70 countries encountered difficulties accessing Microsoft products. The outages affected cybersecurity software, file sharing services and administration tools. The company offers all of these products, along with office applications Word and Excel, as part of the Microsoft 365 suite. Microsoft said last month that the suite has more than 450 million commercial users.

The problem was caused by a patch that was prepared specifically for Microsoft 365 network gateway servers. The rollout began in July, and for more than three months the changes worked worldwide without any visible problems. Microsoft emphasizes that neither the error itself nor this fix affects standard versions of Windows installed on user computers.

A conflict arose within one of the Windows system processes: several tasks simultaneously tried to access shared resources (the so-called “lock contention” or lock contention). With certain volumes of traffic, this led to exhaustion of processor and RAM resources. Connection requests piled up until the servers stopped responding. During testing, Microsoft specialists did not identify the problem, since the tests were carried out on less powerful hardware, while on more modern systems with a large number of cores the error was much more severe.

A combination of two factors led to the error. Approximately 15 hours before the outage, Teams service traffic was redirected from an alternative gateway to the main one, which was also used for Exchange Online. At the same time, about 14% of the gateway server resources were unavailable due to another problem unrelated to this incident. According to Microsoft, the remaining capacity would have been sufficient for normal operation, but the combination of these factors caused the failure.

  Image source: unsplash.com

Image source: unsplash.com

After the servers failed, automated systems redirected connections to healthy nodes. The additional load caused the same error to occur in other locations, causing outages to spread to different regions in less than two hours. At the peak, more than 80% of servers stopped responding. Microsoft’s materials discussing the incident also note that the unavailability of some servers led to additional load on the remaining infrastructure.

A failure in the isolation mechanisms designed to contain problems caused the problem to spread globally. After restarting, the servers soon crashed again due to the renewed flow of requests. While Teams, SharePoint, and OneDrive were able to redirect some of the load, Exchange Online was unable to completely reroute traffic due to the sheer scale of its operation.

Engineers developed an update that returned the connection processing mechanism in Windows to its previous, correctly working state. To do this, they temporarily took large groups of servers out of service, made the necessary changes, performed a reboot, and gradually restored traffic flow. In addition, Microsoft has increased processing power and extended the delivery period for emails so that messages that end up in the queue are not rejected as undeliverable.

The report released by Microsoft is preliminary, with more details on the remediation measures coming on November 20th. In the service status logs, incidents are identified as EX1550543 and MO1550821. According to official Microsoft clarifications, no changes are required to the configuration of tenants or individual devices. The failure occurred exclusively in the Microsoft cloud infrastructure, and not in Windows client systems.

If you notice an error, select it with the mouse and press CTRL+ENTER.