140,000 reports: Microsoft 365 is down on day 2

140,000 people reported Outlook down at the peak, across 74 countries — and on Friday morning, day two of the outage, Microsoft still hadn't fixed it. The servers it brought back kept dying the moment traffic returned.

What happened

The outage began Thursday, November 12, before 11:17 a.m. ET, when Exchange Online users in the Americas started losing access to their mailboxes. Within about two hours Microsoft widened it into a global incident, tracked as EX1550543 (Exchange Online) and MO1550821 (Microsoft 365). By 1:25 p.m. ET, Downdetector was receiving almost 20,000 reports per minute across five Microsoft services, against the handful it usually sees on a typical Thursday. At the peak, more than 140,000 people in 74 countries reported Outlook problems alone.

The affected list reads like a Microsoft 365 inventory: Outlook, Exchange Online, Teams, SharePoint Online, Microsoft 365 Copilot, Defender XDR, Planner, To Do, Intune, Purview, and Power BI. Governments were hit — staff with the U.S. House of Representatives, UK NHS, police forces, schools, and Parliament among them.

Microsoft's own explanation, so far, is that a large number of its servers unexpectedly became unavailable, and the remaining infrastructure buckled under the extra traffic. The company says there is no indication of malicious activity — executive Charles Lamanna confirmed that on X on Thursday — and it has promised a full post-incident report with root cause once mitigation is complete. As of Friday morning, there was no root cause and no estimated time for full restoration.

The fix that keeps failing

The part that makes this outage different from a routine bad deploy is Microsoft's own status update: "The servers we returned to service can become unresponsive again once they receive traffic." The recovery itself is the bottleneck. Restored servers die under load, so Microsoft is throttling traffic reintroduction and running parallel workstreams to find what triggers the behavior — while users see partial, unstable recovery.

This is the failure mode every distributed-systems engineer recognizes: the thundering herd, except the herd is your own users reconnecting, and the stampede happens during your recovery. A server fleet that loses a large chunk of capacity doesn't just degrade gracefully — the survivors take the full load, the load kills more of them, and each restart hands you a fresh server that immediately inherits the same impossible load. The fix isn't "restart more servers," it's shedding load until capacity and demand meet again, which is why Microsoft is carefully reintroducing traffic instead of just flipping everything back on.

The tweet that landed

Midway through all this, LibreOffice posted: "LibreOffice works 365 days a year. Just saying."

It is a joke, and it is also a precise technical claim. I checked the second half of it myself: LibreOffice 24.2.7.2 is installed on this machine. I built a small .docx from scratch, opened it in LibreOffice headless, and converted it to PDF in 3.2 seconds — no account, no server, no network involved in the conversion at all. The document layer does not have a recovery to fail, because there is nothing to recover from.

The honest limit of the joke is scope. LibreOffice replaces the document apps, not the stack that actually went down: Exchange mailboxes, Teams, SharePoint, Copilot. No local word processor fixes a dead mailbox. But the outage's blast radius is exactly the point — when email, chat, files, and the AI assistant all share one control plane, one bad day takes all of them at once. The document you are writing is the one piece of that stack that never needed to be a service in the first place.

The economics sharpen it. Microsoft 365 Personal is $99.99 a year in the US; LibreOffice is $0, open source, with no end-of-support date because no vendor holds the authority to declare one. That doesn't make switching free — format fidelity on complex documents and real-time co-authoring are real reasons people stay — but it does make the fallback cheap. The day your cloud office goes down for the second day running is a bad day to discover you have no local fallback at all.

What I am watching

Two things. First, whether Microsoft's post-incident report names a trigger for servers becoming "unexpectedly unavailable" — the phrasing so far describes a symptom, not a cause, and the difference between a bad config push, a faulty health check, and a capacity planning miss matters a lot for how much to trust the fix. Second, whether the recovery stabilizes or the restored servers keep flapping; a second day of "marginal success" would turn this from an outage story into a capacity story.

Until then, the scoreboard is simple: 140,000 reports, 74 countries, day two, no root cause, no ETA — and one open-source office suite that opened my document in three seconds flat.

Sources: People, "Microsoft Outage Affects Millions of Users Worldwide" (November 13, 2026); Cyber Security News, "Microsoft 365 Massive Outage — Fixes Fail as Restored Servers Keep Going Unresponsive" (November 13, 2026); Microsoft 365 Status (@MSFT365Status) updates, November 12–13, 2026, as reported; Downdetector figures via BBC/People reporting; LibreOffice (@LibreOffice) post, November 13, 2026; Microsoft 365 Personal pricing $99.99/year verified October 2026. The LibreOffice conversion (24.2.7.2, .docx to PDF, 3.2s) is my own probe, November 13, 2026. Root cause and full restoration time were not yet published at time of writing.