Phones & PCs

When the Patch Breaks Something: How to Stop Updates Causing Your Next Outage

← Back to all articles

Most IT outages are not caused by attackers. They are caused by changes: an update, a configuration tweak, a new piece of equipment. November 2026 gave two clear examples from Microsoft alone, and both point to the same answer. The way to avoid the next incident is not to change less, but to change in small, tested, reversible steps, and to plan for the day a provider's change fails anyway.

A months-old bug took Microsoft 365 offline for a day

Microsoft's global Microsoft 365 outage on 12 and 13 November stopped Outlook, Teams and other services for about 24 hours. Microsoft has said it was not a cyberattack. Its preliminary post-incident report, reported by Information Age, sets out three factors:

  • A latent bug from July. An update to Microsoft 365's networking gateway servers carried a defect. It went through Microsoft's safe deployment process and ran worldwide for more than three months without any visible problem.
  • Tests that did not match reality. Microsoft said its load testing used older, less powerful hardware that was less exposed to the bug, so the tests did not catch it.
  • Extra load at the wrong moment. A separate Teams issue shifted traffic onto the gateway that also serves Exchange Online, and about 14% of gateway capacity was already out of service because of an unrelated problem the day before.

Under that load, servers began to stall. Automatic traffic redirection then moved users to the remaining healthy servers, which overloaded them too. Within two hours more than 80% of the gateway servers were unresponsive. Service only stabilised after Microsoft reverted the July change and turned off some of the automatic redirection.

November's Windows updates broke things too

The same month, two problems followed Microsoft's November security updates for Windows, both now fixed:

  • Updates failed to install on some Windows 11 PCs. According to BleepingComputer, KB5132061 failed with errors such as 0x80070306 on PCs that had installed October's optional preview update. Microsoft resolved it on its side within about two days.
  • Activation and licensing commands stopped working. Organisations using the slmgr tool to manage activation, product keys or KMS saw error 0x80041008. Microsoft released emergency updates on 13 November. Devices that were already activated stayed activated.

Lesson one: test on what you actually run

Microsoft's testing missed the bug because the test hardware was not like production. The Windows install failures hit PCs in a particular state, those that had taken a preview update. In both cases, problems surfaced only on systems that looked different from the test environment.

  • Build a pilot group that mirrors your real estate: the same hardware models, the same Windows builds, the same key applications and scripts.
  • Keep optional preview updates on test machines, not across the whole business.
  • Test the things your business depends on, including printers, VPN, line-of-business applications and any automation such as activation scripts.

Lesson two: roll out slowly, watch closely and be ready to reverse

Microsoft's bug sat quietly for three months. Not every problem appears on day one, so a rollout plan needs more than a test.

  • Use update rings. Pilot group first, then a wider group, then everyone, with a few days between each. Windows Update for Business and Intune both support this.
  • Watch for early warning signs. Check Microsoft's Windows release health page before widening a rollout, and watch your own helpdesk tickets and monitoring for unusual patterns.
  • Have a rollback plan for every change. Microsoft's recovery came from reverting the change. Make sure your team knows how to pause, uninstall or revert, and has tested it.
  • Keep a change log. When something breaks weeks later, a record of what changed and when is the fastest route to the cause.

Lesson three: leave headroom, because failures cascade

The Microsoft outage turned serious because spare capacity was already reduced and automatic failover pushed load onto servers that could not take it. The same pattern happens in on-premises networks, storage and virtualisation clusters.

  • Run with spare capacity, so losing one server, link or node does not overload the rest.
  • Know what fails over where. Check that your failover paths can carry the full load, not just part of it.
  • Fix "small" faults quickly. A component left out of service makes the next incident worse.

Lesson four: you cannot control your provider's changes, so plan around them

No amount of testing on your side prevents a cloud provider's bug. What you can control is how long it stops your business. Agree a backup communication channel, keep key contacts and files reachable outside Microsoft 365, and keep an emergency admin account ready. We covered this in detail in our Microsoft 365 business continuity checklist.

Should you just delay updates?

It is tempting after a month like this. But November's Patch Tuesday fixed more than 1,000 vulnerabilities across Microsoft products, and attackers increasingly turn published fixes into working exploits within days. Delaying updates for weeks trades an occasional, fixable glitch for a much larger risk. The answer is a controlled process, not a slower one: aim to have security updates on every device within an agreed time, for example 14 days, and faster for flaws already being exploited.

Our view

Change is unavoidable, and so are occasional bad changes, even at Microsoft's scale. The businesses that come through them best test on systems that look like production, roll out in stages, keep headroom and a way back, and have a plan for when a provider fails anyway.

Microsoft's full post-incident review is still to come. For what Microsoft is planning next across Windows, Microsoft 365, Copilot and security, find what to explore next at Microsoft Ignite, running 17 to 20 November in San Francisco and online.

If you would like help setting up update rings, reviewing capacity and failover in your network and servers, or planning for cloud outages, contact our team.

Drawing on reporting from Information Age (ACS), BleepingComputer, Microsoft.

Chat with us WhatsApp