Microsoft 365's 1,430-Minute Outage Traces to a July Windows Change
Software / news
Microsoft 365's 1,430-Minute Outage Traces to a July Windows Change
Microsoft's preliminary report blames a latent lock-contention defect, a Teams failover and 14 percent of gateway capacity already offline, and customers have until the end of December to claim service credits.

Microsoft's preliminary post-incident report for the Nov. 12 Microsoft 365 outage puts the cause on a Windows change shipped to its networking gateway servers in July, which sat for more than three months before three conditions lined up. Exchange Online customers can claim service credits until the end of December, according to analyst Tony Redmond.
The outage ran from 4:20 PM UTC on Nov. 12 to 4:10 PM UTC on Nov. 13, BleepingComputer reported, citing the report. That is 23 hours and 50 minutes, or 1,430 minutes. Microsoft tracked it as incident EX1550543 for Exchange Online and MO1550821 for knock-on effects, Redmond wrote on Office 365 for IT Pros.

Three conditions, none enough alone
Microsoft attributes the outage to three factors, which Charles Lamanna, its executive vice president for Copilot, Agents and Platform, described in an update on X, ITPro reported.
| Factor | What the report says |
|---|---|
| Latent defect | Lock contention in a change deployed from July overloaded gateway servers |
| Load surge | Teams failed over to the gateway Exchange Online uses, about 15 hours earlier |
| Reduced capacity | About 14 percent of gateway capacity was out for an unrelated reason |
Microsoft said the infrastructure had enough headroom for the 14 percent loss alone. Redmond's summary adds that the defect escaped testing because contention is likelier on newer, higher-core hardware, while Microsoft's lab testing used weaker machines.
Within two hours, traffic redirection turned a local fault into a cascade. Redmond quotes Lamanna describing "cascading failures + thundering herds at our network perimeter". A separate problem let the failure cross the isolation boundaries between gateway capacity units and regions, and Exchange Online could not fail away from the gateway at the scale needed, which is why it was hit longest. Microsoft said Teams, OneDrive and SharePoint could move to other infrastructure and recovered earlier.
Why a rollback did not work
The defective change had shipped more than three months earlier, so ordinary rollback was no help. Microsoft developed a fix that restored the previous known-good Windows code path, Redmond wrote. Servers were taken offline in large groups, patched, restarted and returned to service gradually. Recovered servers were first swamped by backed-up traffic and failed again.
Lamanna said there was no indication of malicious or cyber activity.
What the credit is worth
Redmond's arithmetic: November has 43,200 minutes, so if every user lost access for all 1,430 minutes, monthly uptime would be about 96.69 percent. That lands in the 95 to 99 percent band he reads as a 50 percent service credit. The figure is his estimate for a worst case, since Microsoft measures uptime in user-minutes and not every user was affected for the full window.
Exchange Online Plan 2 lists at $8 per user per month in the United States, Redmond said, and Microsoft 365 E5 at $60. Credits are not automatic: admins file a claim through the admin center or their reseller by the end of December, and the credit is the sole remedy, applied against future fees. For comparison, Redmond put Microsoft 365 uptime at 99.526 percent in the first quarter of 2026, his worst since 2013.
Related reporting: the Windows Server 2028 hardened Server Core build and Synology's third critical DSM advisory in 62 days show how patch pipelines carry their own failure modes.
Microsoft has said a final post-incident report with longer-term fixes will follow. None of the three sources read for this story contained it.
Sources
More in Software
- 01Brazil's Receita Federal Posts Tax Calculator Under MIT With 16 Open IssuesThe agency published three repositories on Nov. 16 for the new consumption taxes, and outside developers had filed bug reports against the first within days.
- 02OpenSRE Hits 13,884 Stars After Repositioning Toward CI/CD FailuresThe Apache-2.0 agent is labelled public alpha, ships a dated release almost every day, and sends analytics and crash reports unless users set an environment variable.
- 03Firecrawl's Alexandria Data Library Is Missing From Its 205,955-Star RepoThe AGPL-3.0 repository covers scraping and a self-hosted server, while the 550-provider data library and search run on Firecrawl Cloud.
- 04Vim Classic 8.3.1 Backports Patches to 8.2.3237 and Drops Python 2The AI-free Vim fork's second release is signed by Drew DeVault, and its announcement does not say which CVEs the new patches fix.