The week in brief
The sequence below runs from 12 to 17 November 2026. Treat the clustering as a coincidence unless evidence says otherwise.
- Microsoft 365 (12–13 November): The incident IDs are EX1550543 for Exchange Online and MO1550821 for the wider Microsoft 365 service. Microsoft's preliminary post-incident report puts the window at 16:20 UTC on 12 November to 16:10 UTC on 13 November, which is 23 hours and 50 minutes.
- FAA (13 November): The FAA reported a telecommunications problem that led to a nationwide ground stop of more than four hours. Its own account says legacy equipment became unstable when the network was flooded with connection attempts. The FAA page itself wasn't retrievable for this piece, so these details come from reporting on the FAA's explanation. Some coverage also mentions contractor maintenance work, so the full cause may be more complicated than "old kit got overwhelmed."
- AWS (15 November): The CITSC status page, which tracks vendor incidents, records signup, authentication and connectivity errors in US-EAST-1. That page says authentication was restored at 8:20 AM ET and the remaining load balancer issue was fixed at 9:21 AM ET. AWS had not published a cause. This was a regional problem affecting particular services, not an AWS-wide shutdown.
- Apple (17 November): Apple listed outages across iCloud sign-in, the App Store, iCloud Mail and other services. Apple says the cause was an internal update and not a cyberattack. That is Apple's explanation, not an independently confirmed root cause.
- JPMorgan Chase: The original commentary says a software problem delayed transactions, including some involving other banks' customers. I couldn't verify that detail independently, so it should be read as a claim, not a confirmed banking-system outage.
What went wrong at Microsoft
This is the part of the week that matters most to Windows and Microsoft 365 administrators.
According to Microsoft's preliminary report, as summarised by Charles Lamanna, Microsoft's EVP of Copilot, Agents and Platform, three main factors together caused a global outage of the networking gateway that fronts many Microsoft 365 services.
- A latent defect. Servers became overwhelmed and unresponsive under certain load conditions. Reporting dates the underlying change to July 2026.
- A Teams failover. About fifteen hours before the outage, Teams failed over from the alternative gateway to the primary gateway that also supports Exchange Online. Microsoft says this kind of failover is a normal part of operations and of its resiliency design.
- Reduced capacity. Nearly 14% of the gateway's server capacity was out of service because of an unrelated issue the previous day.
Microsoft's incident updates show how the failure spread. The updates described servers becoming intermittently unavailable, which put extra strain on the remaining infrastructure. Microsoft redistributed Outlook connection traffic across its wider infrastructure to balance the load. In a separate status roundup, Microsoft's report was said to put the lowest point at more than 80% of gateway servers unresponsive. That 80% figure comes from a secondary status page, not from Microsoft directly.
Because the defect had been in place since July, there was no recent change to roll back. That fits the original commentary's point that a simple revert wasn't available.
Not everything failed equally
Exchange Online took the worst of it. Reporting indicates Teams chats and meetings worked for most of the outage, and some services could move away from the faulty gateway and recover earlier. Services that depend on Exchange, such as Teams calendar features and Copilot responses that draw on Outlook data, were affected.
Microsoft's incident records also list wider impact. A University of Pennsylvania mirror of the incident text lists Purview services, including data loss prevention, as affected. Don't read that as every service being fully down for the whole period. Impact varied by service and by region.
Windows PCs themselves were not the problem. One analysis states that Windows PCs were unaffected. Cached mail in classic Outlook could still be read.
Final report
The preliminary report is not the final word. Coverage says the final report was due on 20 November 2026, and admins should look up the incident IDs and compare the two. I haven't seen the final version, so treat the root-cause details as provisional.
Why small faults spread
The Microsoft case is a textbook example of a resilience mechanism turning into a transmission path. Failover is meant to move traffic from a sick component to a healthy one. If the healthy one shares the same latent flaw and is already short on capacity, the extra load can tip it over too. That is general engineering reasoning, not a quote from Microsoft, but the incident updates match it.
Precedents
Cloudflare, 18 November 2025
Cloudflare's own postmortem says the outage wasn't caused by a cyberattack. A permissions change led a database to output duplicate entries into a Bot Management "feature file," which doubled in size. That oversized file was pushed to machines across the network. The routing software had a limit below the doubled size, so it failed. Cloudflare says core traffic was largely flowing normally by 14:30 UTC. All systems were normal by 17:06 UTC.
The original commentary says this happened "almost exactly a year" before the November 2026 incidents. More precisely, it was one year before 18 November 2026, so it falls just after the week in question.
CrowdStrike, July 2024
A faulty CrowdStrike update caused serious disruption to Windows machines. Microsoft estimated about 8.5 million devices were affected, which was less than 1% of all Windows devices. The trigger was CrowdStrike's update, not a Microsoft update, and the affected machines ran Windows.
The useful lesson is that device count is a poor measure of impact. A few machines running airport, hospital or banking systems matter far more than the same number sitting idle.
Practical steps
These are general preparedness steps, not a guarantee against any given outage.
For individuals
- Keep local copies of important documents and photos.
- Download travel documents and directions before you need them.
- Keep an alternative way to pay, including a little cash.
For IT teams
- Check whether your fallback email or chat path shares the same cloud, network, identity or authentication dependencies as the primary system. A backup that depends on the same provider isn't independent.
- Know where to find incident IDs and post-incident reports in the Microsoft 365 admin center.
- Test what happens when a key supplier goes down, not just when a single server does.
- During an outage, be wary of "Microsoft outage" emails that ask you to sign in. One guide advises against resetting passwords or reinstalling Office as a response.
Is technology getting less reliable?
One bad week doesn't prove it. These systems are more complex and more widely used than ever, so individual failures can feel bigger even if overall reliability is improving. One analysis says per-site outages are declining, which supports caution here.
A narrower point holds up well. As more organisations lean on a handful of providers, correlated failures cost more when they happen. Whether that justifies more regulatory attention is a judgment call. The original commentary claims a UK minister told the House of Commons that over 90% of the civil service uses a Microsoft product. I couldn't verify that, so I've left it out as a fact.
The takeaway
The failures were ordinary: software defects, traffic surges, an internal update and capacity already lost. They weren't the work of elite hackers. The question for IT teams is less "will this happen?" than "what do we do when it does?" If your answer is "wait for the provider," it's worth testing a better one before the next bad week.
References
- Why Does the Internet Keep Breaking? - Mark Kelly | Substack Mark Kelly | Substack · 2026-11-21T11:40:18+00:00
- Microsoft Outage November 2026: Causes & Latest Updates recoveryfix.com
- Microsoft 365 Outage Nov 2026: Resolved, Cause, Next Steps epifive.com