Watch the video, transcript and sources
Friday, November 13, 2026 · The Daily Diff · Reporting cutoff: Microsoft's 11:04:50 UTC update
https://www.youtube.com/watch?v=R5v9I-_1uFE
- Returned mail servers could fail again when traffic returned
+ Most core Teams functionality and direct SharePoint files remained functional
+ A longer mail timeout aimed to preserve delayed messagesMicrosoft returned mail servers to service. Its own updates then described them becoming unresponsive again when traffic arrived. For developers, the useful follow-up is understanding which operations depended on the impaired service and what a temporary return actually proved.
Start with the dependency behind the operation. Exchange Online is Microsoft's hosted email service. Microsoft's expanded service-health notice said most core Teams calls, meetings and chats remained functional. Calendar and scheduling scenarios depended on Exchange and could fail. Direct SharePoint file access remained functional, while some sharing and Exchange-dependent features were affected.
That explains how a meeting can proceed while the calendar refuses to cooperate. The product logo is a poor map of the backend needed for your next click.
Capacity disappeared
Microsoft initially investigated Exchange mailbox access in the Americas on Thursday, then confirmed that users in any region could be affected. CNN independently reported disruption across banks, governments and courts. Its reports of affected countries came from Downdetector reports; the episode makes no claim that entire countries or every Microsoft service stopped working.
The official account attributed elevated load to unexpectedly unavailable servers straining the remaining infrastructure. Ordinary demand can overwhelm a smaller available pool. The company investigated automated health checks and a component responsible for processing server health. That explanation remained an investigation hypothesis at the reporting cutoff.
Microsoft paused deployments, redistributed traffic and restarted affected servers. An alternative health-check method brought a subset back. Automated systems were being prepared to detect and restart unresponsive machines. These updates described recovery attempts, with results still being tested.
Traffic came back too
The most useful primary admission was specific: returned servers could become unresponsive again once they received traffic. Microsoft also described isolating infrastructure, restarting components and reintroducing traffic to confirm stability before applying changes more broadly.
The episode follows the same conceptual server through unavailable, returned, taking work and unresponsive states. That illustration communicates the reported sequence. Microsoft's final report still needed to establish the initial trigger and the conditions behind the relapse.
Microsoft limited certain traffic reaching affected infrastructure in some locations and rebuilt a subset of servers using a previous configuration to test stability. A deployment pause during recovery does not establish that a Friday deployment caused the original incident. The public notice also said Microsoft had no evidence of malicious activity and would publish a full incident report after mitigation.
By the official Friday update at 11:04:50 UTC, recovery actions had completed across portions of affected infrastructure in Asia Pacific. Traffic was gradually returning, with improvements in mailbox access and delayed email processing. Some users could still experience intermittent issues, and recovery in other regions continued. That is the episode's reporting cutoff. The earlier 10:01 update described variable health and capacity below baseline; it remains part of the recovery history.
Check the actual request
Microsoft Graph requests for Exchange mail, calendar and contacts could time out. An unrelated successful operation therefore offered little assurance about a mailbox operation's outcome. Review the specific dependency and the result the user needs.
Microsoft also published incident information outside its admin center because administrators might be unable to load service-health pages there. Keep an out-of-band status link available through an approved incident process. A dashboard depending on the impaired service has become part of the incident.
For ordinary Graph throttling, Microsoft's documentation recommends honoring Retry-After and using exponential backoff when the response lacks that header. The documentation concerns throttling behavior. It supplies no diagnosis of this particular outage and no guarantee that retries restore an unavailable service.
What the forty-eight hours meant
Microsoft's expanded notice said it prioritized queue and backlog processing and extended the email timeout window to forty-eight hours to preserve delayed messages during recovery. That gives delayed mail more time. The notice supplied no forty-eight-hour repair deadline or guarantee that every message would arrive.
A delayed notification has not necessarily disappeared. A queued message has not necessarily reached its recipient. For time-sensitive work, verify the actual outcome through an approved independent channel before relying on an optimistic client window.
Verdict: NEEDS REVIEW
I'd keep recovery under review because Microsoft's updates described returned servers failing under traffic. Sustained service with real work coming through is the useful test. At the reporting cutoff, the initial cause still awaited the final incident report.
And that's the diff for today. I'm Niko from Axrisi. Merge responsibly.
Sources
Expanded service-health incident — Microsoft; reporting cutoff November13,11:04:50UTC, with earlier10:01:07observations retained historically.
Initial Americas investigation and possible impact in any region.
Visual credits
Satya Nadella portrait by OFFICIAL LEWEB PHOTOS (2013), CC BY2.0, Commons source. Whole portrait-based artwork is adapted through cropping, background removal, resizing and text/composition. Vendor identity only; no endorsement, report authorship or current reaction is implied.
The WWE still and The Simpsons/Bob rake GIF are brief silent process commentary. WWE, The Simpsons/20th Television/Disney, Microsoft and other source owners retain copyright. Hosting and this specific editorial-use decision provide no blanket reuse grant or copyright clearance. The portrait licence is separate from third-party source/cartoon rights.
Visual credits: Satya Nadella portrait by OFFICIAL LEWEB PHOTOS (2013), CC BY 2.0: https://commons.wikimedia.org/wiki/File:Satya_Nadella.jpg and https://creativecommons.org/licenses/by/2.0/ . Whole portrait-based artwork adapted by cropping, background removal, resizing and text/composition; vendor identity only, no endorsement or current reaction. WWE still: https://imgflip.com/memetemplate/234202281/AJ-Styles--Undertaker . The Simpsons/Bob GIF: https://giphy.com/gifs/the-simpsons-sideshow-bob-RSOUOj8H9A3Xq . Brief silent process commentary; WWE, The Simpsons/20th Television/Disney, Microsoft and other source owners retain copyright. Hosting and this specific editorial-use decision provide no blanket reuse grant or copyright clearance. The portrait licence is separate from third-party source/cartoon rights.