Microsoft 365 Outage, Hour by Hour: What the Preliminary Report Reveals

Cloud of servers with crashed equipment and disconnected mail and chat icons

The preliminary report Microsoft sent to its customers about the Microsoft 365 outage makes it possible to reconstruct, almost hour by hour, how an overnight traffic shift combined with a defect that had been lurking since July to leave most Exchange Online users without email for roughly 24 hours, between 16:20 UTC on November 12 and 16:10 UTC on November 13, 2026. The document, dated November 15, can be downloaded at the end of this article.

The Microsoft 365 outage timeline in 20 seconds

  • 01:07 UTC on 11/12: Teams shifts to the main gateway it shares with Exchange Online.
  • 16:20 UTC: impact begins; in under two hours, more than 80% of the servers go down.
  • 07:20 UTC on 11/13: the mitigation that rolls back the code begins.
  • 16:10 UTC: Microsoft declares the incident resolved; the final report is due on 11/20.

We already covered the cause Microsoft attributes to the outage: a lock-contention defect in an internal Windows process, introduced by a fix meant only for Microsoft 365’s network gateway servers. With the full document — tracked under incidents EX1550543 and MO1550821 — it’s possible to lay out the sequence of events and see which services went down, when each region recovered, and which remediation steps Microsoft now considers complete.

From July to November 12: three factors line up

The story starts months earlier. In July, Microsoft began rolling out a Windows fix across its network gateways that was custom-built for that infrastructure and, according to the report, does not affect general operating-system installations. It ran for more than three months with no visible impact. Load testing missed the problem because it ran on less powerful hardware, with fewer cores, than production — where the defect turned out to be worse.

On November 11, an unrelated problem knocked out nearly 14% of gateway capacity. Microsoft maintains the remaining capacity would have been enough on its own. At 01:07 UTC on November 12, roughly 15 hours before the outage, Teams traffic shifted from an alternate gateway to the main one — the same gateway Exchange Online uses. With more load on fewer servers, the gateways hit the exact conditions that triggered the defect.

The 24 hours of the incident

Impact began at 16:20 UTC on November 12 — 17:20 in mainland Spain. Within two hours the problem had spread, and at its worst more than 80% of gateway servers weren’t responding. The barriers meant to isolate partitions from one another didn’t hold, and automatic recovery — by shifting connections onto healthy servers — overloaded those servers too and took them down as well.

Exchange Online was the main service hit: most users couldn’t reach their mailbox, and mail arrived late. Because of its scale, Exchange couldn’t move its traffic off those gateways. The report also lists a long string of services with secondary impact: Microsoft 365 Copilot, Teams, SharePoint, OneDrive, Defender, Purview, Fabric, Intune, Planner, Power Apps, To Do, Power BI, and Universal Print.

The usual tools for undoing a change — rolling back a deployment, staged rollout (flighting), or feature flags — didn’t work for a change applied months earlier. Engineers had to prepare an update that would revert the code to its previous state. At 07:20 UTC on November 13, that mitigation began rolling out: pulling groups of servers out of service, updating them, and bringing them back online gradually.

According to the report, Europe had recovered by 11:35 UTC. Web-mail access was restored globally at 14:17 UTC, and Microsoft declared the incident resolved at 16:10 UTC. The last customer updates went out on November 14 at 04:15 UTC, as queued mail continued draining.

What Microsoft considers done, and what’s still open

The document lists several measures as complete: the update that reverts the change, an extended mail-delivery window so queued messages weren’t bounced as undeliverable, additional gateway capacity, more monitoring, and new safeguards in how traffic gets distributed. Microsoft hasn’t yet detailed what will change in its load testing, in partition isolation, or in how it rolls back older changes. The company points to the final version of the report, due November 20, 2026.

For businesses, the timeline has a practical use. The global window Microsoft gives is a reference point, but the hours that actually count for a service-credit claim are the ones shown in each tenant’s own Service Health dashboard, which can vary by region. We explained the process for claiming Microsoft’s SLA credit over the outage separately.

The original preliminary report can be downloaded here: Microsoft’s preliminary report on the November 2026 Microsoft 365 outage (PDF, 121 KB).

Frequently asked questions

What time did the Microsoft 365 outage start?

At 16:20 UTC on November 12, 2026 — 17:20 in mainland Spain — according to the preliminary report.

When did Europe recover?

Microsoft puts Europe’s recovery at 11:35 UTC on November 13, ahead of the global restoration of webmail access at 14:17 UTC.

Which services were affected?

Mainly Exchange Online, with secondary impact on Copilot, Teams, SharePoint, OneDrive, Defender, Purview, Fabric, Intune, Planner, Power Apps, To Do, Power BI, and Universal Print.

When will Microsoft publish the final report?

It’s due on November 20, 2026, with the full list of remediation measures.

Sources:

This article is also available in Spanish: “Caída de Microsoft 365 hora a hora: lo que revela el informe preliminar,” on revistacloud.com.

Scroll to Top