System status
What is running right now, and everything that has gone wrong before.
We could not reach the status service
Past incidents
Every outage we know of, in plain words. A percentage of uptime is not published yet: we will show one once probe results have been stored long enough to compute it honestly.
Application server stopped responding
What you would have seen. The whole application was unavailable while the static site kept loading. Signed-in work could not be opened or saved.
How long. Hours; no probe was watching the server, so it was found by hand.
Why. A pattern-matching routine in a background job that reads a government source took exponentially long on one page and held the server's single thread.
The heavy background jobs run in a separate, lower-priority process, so they cannot starve the application.
A health probe that needs the server to be responsive restarts the container automatically when it fails.
Pattern matching on untrusted text runs in an isolated worker with a time limit, and a check fails the build if new code bypasses it.
Production deploys wait for a health check and roll back on failure.