Sluether :v_aroace: :wolfmoon:SuperSluether@tech.lgbt
Nov 16, 2026, 5:30 PMThe post incident report for last week’s 365 outage is fascinating. They had a bug that made it past testing in July where servers could run out of resources during high traffic, but it’s worse on higher end servers with more cores and their testing focuses on servers with fewer cores.
That, combined with a typical Teams automated failover, plus a 14% capacity reduction due to planned maintenance lead to cascading failures. Servers got overloaded and stopped responding, automated recovery and load balancing did the rest until 80% of capacity was unresponsive.
And since the bug was introduced in July, it had already been deployed globally and wasn’t a recent change that could easily be pinpointed.