
Last week’s Microsoft outage, which cut businesses around the world off from their email, can be traced to a fix the company rolled out in July, according to a preliminary post-incident report Microsoft has sent to customers. Its own failover systems then spread the failure worldwide.
The report, dated Sunday, puts the length of the outage at almost 24 hours, from 6.20pm South African time on Thursday, 12 November, to 6.10pm on Friday.
The worst-hit service was Exchange Online, the email system behind Outlook. Most users couldn’t get into their mailboxes, and email delivery was delayed. In South Africa, the South African Revenue Service and the City of Cape Town were among the organisations that warned users of disruption.
The July fix
The faulty change was written specifically for the servers in a networking gateway that sits in front of many Microsoft 365 services, receiving users’ connections before passing them on to services such as Exchange. Under certain loads, it caused tasks inside an internal Windows process on those servers to get stuck waiting on one another, until the machines ran out of processing power and memory and stopped responding.
The change went through Microsoft’s staged deployment process from July. “For more than three months, the change was deployed worldwide without impact,” the report says. Microsoft’s load testing missed the fault because it is worse on newer machines with more processor cores, while the tests “tended to focus on weaker hardware with fewer cores”. The defect does not affect other Windows deployments, Microsoft said.
This caused a cascading failure from server to server as the same amount of load was spread across fewer servers.
Microsoft
At 3.07am on Thursday, Microsoft’s systems automatically moved Teams traffic from a second, independent gateway onto the one that also carries Exchange Online, which Microsoft describes as a normal part of operations. Nearly 14% of that gateway’s server capacity was already out of action because of an unrelated problem the day before, which Microsoft says had no impact on its own.
Together, the extra load and the reduced capacity triggered the bug during peak US traffic. Isolated server failures had already begun in some regions during the day, and automated fixes were applied, the report shows.
How it spread
When servers stop responding, Microsoft’s traffic management systems send users to healthy servers instead. On Thursday, the redirected load pushed those servers into the same fault.
“This caused a cascading failure from server to server as the same amount of load was spread across fewer servers,” Microsoft said. In less than two hours, the fault had spread across the gateway’s servers, and at the low point more than 80% of them were unresponsive.
The gateway is divided into thousands of units grouped into regions, with barriers meant to stop a failure in one spreading to the rest. They did not hold, “due to a separate issue” that the preliminary report does not explain.

Servers brought back online were immediately swamped by waiting traffic and became unresponsive again. Microsoft said its automated recovery processes “were not designed to restore service from this broad unresponsive state”. Because the faulty code was months old, its usual tools for rolling back changes could not quickly undo it.
Engineers tried new health checks, restarts, traffic blocking and extra capacity through the night. At 9.20am South African time on Friday, 15 hours after the first alert, Microsoft began applying an update that reverted the faulty change. It took groups of servers out of service, applied the update and restarted them, prioritising regions in or entering business hours. Europe was back to normal by 1.35pm, and webmail worldwide by 4.17pm.
Why email took longest
Teams, OneDrive, SharePoint and the Office apps were able to switch away from the failing gateway, so they partly recovered sooner. Microsoft says Teams chats and meetings worked “for most of the outage”, though the City of Cape Town reported intermittent Teams connections on Friday. Exchange Online could not switch, “for scale reasons”, Microsoft said.
Since the outage, Microsoft has added monitoring of how traffic is redirected, put limits on how load is redistributed to avoid a repeat cascade and added capacity. It also extended its email delivery window so that messages queued during the outage were delivered rather than bounced back as undeliverable.

What customers can claim
Microsoft’s service level agreement for its online services, in its October 2026 version, pays customers service credits if Exchange Online falls short of set targets. The main availability measure counts downtime as time users cannot send or receive email using Outlook on the web, the browser version of Outlook. Uptime is calculated across the calendar month and all of a customer’s users. Below 99.9%, the credit is 25% of the fees for the affected service. Below 99%, it is 50%, and below 95%, it is 100%.
In November, 99% allows for just over seven hours of downtime and 95% for 36 hours. Microsoft’s timeline puts webmail back about 22 hours after its first alert, so a customer whose users were cut off throughout would qualify for a 50% credit on that measure.
The agreement has separate targets for email delivery times and undelivered email, and customers must choose one target per incident.
Credits are capped at the month’s fees and are customers’ ‘sole and exclusive remedy’, and they do not cover lost revenue or other losses.
Credits are not automatic. Administrators must lodge a claim with Microsoft support by 31 December, with details of the times, affected users and errors. Credits are capped at the month’s fees and are customers’ “sole and exclusive remedy”, and they do not cover lost revenue or other losses.
Microsoft says it will deliver a final report, with a full list of repairs, by 20 November. – (c) 2026 NewsCentral Media





