Status

Every ~5 minutes a scheduled GitHub Actions workflow requests each service's public address (the dashboard, /v1/auth/options on the API, the proxy's /healthz, this website and the docs) with a 10-second timeout and one retry, and records whether it answered and how long it took. GitHub sometimes runs scheduled jobs late, so gaps of a few extra minutes are normal.

When a service fails two runs in a row, an incident issue is opened on GitHub; it is closed when the service answers again. The results are public as status.json, and the checker is in the repository. Checks run from GitHub's network, so a problem only some regions see may not show here.