**GitHub's reliability crisis 🧵**
May '25–Apr '26: 257 incidents, 48 major — ~1 serious outage/week. Feb '26 the worst ever (37). Actions, the CI/CD backbone, took 57 hits.
CTO blames "agentic workflows" (Copilot & co.). The catch: one PR hits Git storage, Actions, search, webhooks, APIs & DBs at once — a load pattern built for human, not machine, traffic.
But the "10-14x growth" charts shipped with no Y axis. Real data: ~3.5x over 2 years. A chart without scale sells a feeling, not a fact.
Now migrating off maxed-out bare-metal datacenters to Azure (~40% by May, up from 8%). Hard part: bare-metal MySQL doesn't move with a click.
The real question isn't load. It's whether they finish the migration without breaking the service.