At around 4:00pm GMT on Thursday, November 12, Outlook stopped answering. Within an hour, Downdetector was logging reports from 74 countries, and for the next day and a half, a large slice of the working world rediscovered what life looks like without Microsoft 365. For millions of workers, it was a holiday. For MSPs, it meant a long night of tickets nobody could fix—and an unplanned stress test of how much of their clients’ businesses (and their own) now hangs on a single tenant.
What actually happened with Microsoft 365
Microsoft first acknowledged problems in the Americas at roughly 5:54pm UTC, logging an Exchange Online mailbox connectivity incident (EX1550543) before opening a broader multi-service incident (MO1550821) as Teams, Copilot, and OneDrive began to wobble. A separate, shorter Teams incident had already hit the Americas that morning, which muddied early triage.
The impact was massive. In the UK, the National Health Service lost email and Teams for the day, and government bodies and businesses around the world were affected. Many workers simply closed their laptops, and millions of dollars in productivity were lost.
The mechanics look like a textbook cascading failure. Microsoft said servers failed automated health checks and dropped out of service, pushing load onto the servers still standing—and stressed that the extra load wasn’t coming from increased traffic. The survivors then buckled too. The eventual explanation was a defect introduced during an earlier network capacity upgrade, which sat latent until a later network change exposed it. Engineers rolled the change back, but recovery was stubborn: restored servers fell over again as clients reconnected en masse, so Microsoft brought capacity back gradually, validating stability at each stage. Full service, including cleared mail backlogs, wasn’t confirmed until the morning of November 14.
Microsoft has ruled out a cyberattack (and, pointedly, AI), apologized, and promised a full post-incident review. It follows two other major Microsoft 365 incidents this year: a roughly nine-hour North American outage in January, blamed on capacity constraints during maintenance; and an August outage in which a faulty core authentication configuration took down Exchange Online, Teams, SharePoint, and other office-y tools.
The lessons hiding in the root cause
The scariest detail is the latent fault. A change went in, passed whatever validation it faced, and only blew up when a second change landed on top of it. Anyone pushing patches, scripts, and policies across hundreds of tenants through an RMM platform will recognize that risk.
The health-check cascade carries its own warning. Automated remediation designed to protect a system can amplify a failure when nobody has modeled what happens if many nodes fail simultaneously. Your own auto-healing scripts, failover rules, and alert-driven workflows deserve the same scrutiny.
Resilience checks MSPs should run now
Map the real blast radius
Entra ID, Exchange Online, Teams, SharePoint, and Intune tend to be treated as separate services, but they’re tightly interdependent: almost everything authenticates through Entra ID, and Teams leans on Exchange for calendars and meetings and on SharePoint and OneDrive for files and recordings.
Build an out-of-band comms path
Microsoft pointed admins to an external status page in case the admin center itself wasn’t reachable. You need the equivalent: SMS or a status page hosted outside Microsoft’s Cloud, plus an internal channel for your techs that doesn’t run on Teams.
Decide your mail continuity stance
Sending servers typically retry for a day or more, so inbound mail usually arrives late rather than vanishing. For clients in legal, healthcare, or finance, though, late may not be good enough. Third-party continuity and spooling services deserve a proper conversation, priced against the cost of a lost day.
Stage your own changes
Ring-based deployment, pilot tenants (starting with your own), and a clear rollback or pause plan for RMM pushes—with a defined fast-track for genuine emergencies. If Microsoft can ship a latent defect, so can you.
Reread your contracts
Microsoft’s financially backed SLA pays service credits against license fees, not compensation for lost revenue. Make sure your own SLAs are explicit about upstream provider outages, and that clients understand the difference before the next one.
Outage lessons apply to attacks
This incident wasn’t malicious, but the operational experience—core systems dark, comms down, clients anxious—closely resembles what a serious cyber incident delivers, only with a quicker recovery. The 2025 attacks on M&S, Co-op, and Jaguar Land Rover showed how long the darkness lasts when the cause is hostile, and how often attackers get in through IT help desks and third-party suppliers. MSPs, holding privileged access to dozens of client environments, remain a prime target; Kaseya in 2021 proved the model.
So run the tabletop exercise both ways: “Microsoft 365 is gone for 36 hours” and “our RMM has been compromised.” Check that break-glass accounts work and are excluded from conditional access policies that could lock them out. Confirm M365 backups are immutable and independent of the tenant they protect. Backup won’t help with an outage like last week’s, but when ransomware is involved, it’s the difference between a bad week and a business-ending one.
Regulators are moving the same way. NIS2 brings managed service providers into scope across the EU, DORA puts ICT third-party concentration risk squarely in front of financial firms, and the UK’s Cyber Security and Resilience Bill extends oversight to MSPs. Resilience is fast becoming something clients will ask you to show evidence for, not just promise.
Turning disruption into a conversation
The upside is that every client felt this one, which makes it the easiest resilience conversation you’ll have all year. Bring them a dependency map, an honest view of what would and wouldn’t keep working, and a costed set of options. The MSPs who do that now will look very different from the ones still saying “it’s a Microsoft problem” when the next outage hits.
Want to compare notes with MSPs, vendors, and Cloud providers working through exactly these questions? Join the conversation at MSP GLOBAL 2027. Click here for a free pass.




