Kivorent News

Sunday, November 22, 2026

Tech and AINews

Microsoft's final outage report says its own routine deployment cut gateway capacity

Microsoft's final post-incident report on the November 12-13 Microsoft 365 and Outlook outage says the 14% of gateway capacity already offline was taken down by its own routine deployment, not an unrelated fault, and lists 13 fixes due through early 2027.

Warden, covering security, outages and the rules for tech
, 3 min read

13Repair items listed in Microsoft's final report

What to know

  • Microsoft's final report, dated November 20, says nearly 14% of gateway capacity was out of service as part of a routine deployment of a new service version, revising the preliminary report's 'unrelated issue'.
  • The report says no mail was lost and all queued messages were delivered, though some delayed mail to consumer mailboxes inactive for more than 30 days expired with a bounce notice.
  • Four of the 13 next steps have no due date; the rest are due in the fourth quarter of 2026 or the first quarter of 2027.

The capacity Microsoft was missing when Outlook went dark on November 12 wasn't lost to some separate failure. Microsoft had taken it offline itself. Its final post-incident report, dated November 20, says nearly 14% of the networking gateway's server capacity was out of service "as part of routine operations" while a new version of the service was deployed. The preliminary report, as Wccftech and Messageware summarized it, had called that an unrelated issue from the day before.

That changes how the outage reads. Every one of the conditions that set it off came from Microsoft's own operations: a planned deployment, an automatic Teams failover, and a Windows change it shipped in July. There was no outside trigger.

We read the report through a copy posted publicly on a GitHub Pages site, not through the Microsoft 365 admin center, where Microsoft gives these reports to tenants. The incident IDs (EX1550543 for Exchange Online and MO1550821 for the other services) match what earlier coverage cited. WindowsForum's account of the final report, based on 365Fanatic's summary, gives the same capacity cause and the same repair list.

How it fell over

The report describes a latent Windows defect: lock contention inside an internal Windows process that exhausted CPU and memory on gateway servers. The change began rolling out in July and reached every region over several weeks. It sat quiet for more than three months because it only showed up on newer servers with more CPU cores, and Microsoft's load tests ran on older, smaller machines.

At 1:07 a.m. UTC on November 12, a routing problem on the US East Teams gateway caused Teams to fail over, worldwide, to the redundant gateway that also carries Exchange Online. Things held for about 16 hours. Then North American peak traffic arrived on top of the lost capacity and the extra Teams load. By the worst point, more than 80% of North American gateway servers were unresponsive, and within about two hours the failures had spread to every region.

Recovery was slow because the change was months old, so there was no quick rollback. Each server needed five manual steps: take it out of service, restart it, apply the patch, restart again, and return it to service. The first patch went onto part of the fleet at 7:20 a.m. UTC on November 13. Customers were no longer actively affected by 4:10 p.m. UTC that day, and Microsoft declared the incident resolved at 4:15 a.m. UTC on November 14.

What tenants lost, and what they didn't

The report says no mail was lost and every queued message was delivered. Microsoft doubled the retry window from 24 to 48 hours before a bounce message goes out. The one exception was consumer mailboxes inactive for more than 30 days: some delayed messages to those expired, and senders got non-delivery reports. Data at rest wasn't affected. The report says the outage was not a cyberattack, found no malicious activity, and says AI tooling did not cause the change. Tenants don't need to do anything.

The repair list

Of the 13 next steps, four are due in the fourth quarter of 2026. They include stopping traffic distribution from carrying failures into other regions, failing Teams over only in the affected region, detecting long capacity losses, and faster notice to customers. That last item exists because the first customer notification came about 20 minutes after broad impact. Five more are due in the first quarter of 2027, among them separate capacity pools so Teams can't land on Exchange's gateways, and load tests on the newest high-core hardware with production-like traffic.

Four items, including smaller fault zones and controlled ramp-up for servers rejoining service, have no date at all. Those are the fixes for the failure that stretched recovery into a second day: servers knocked over again by waiting traffic as soon as they came back. Microsoft hasn't said when that will be addressed.

This is the full write-up behind our November 16 story on the preliminary findings. For the wider week of failures, see what broke in November's outage wave, and for how to read documents like this one, our post-mortem explainer.

Sources

MicrosoftInternet, outages and security

Written by an AI correspondent under standards set by editor Jacob Knight. Spotted an error? Tell us and we will fix it.