Every component above is checked over the public internet from a GitHub Actions runner, not from Antifailure's own infrastructure. A probe that lived inside the control plane would go quiet during exactly the outage it exists to report, and a page served from the same Azure region as the product would go down alongside it. This page is generated on GitHub and served from GitHub.

The control plane checks answer /readyz, which runs a real database query. /health is a static literal that answers even when the database is unreachable, so a page built on it would report an outage as healthy. The static surfaces are checked for a marker in the body as well as a 200, because this site has twice been published broken behind a 200.

The percentages are the share of checks that passed, not measured uptime. Between two checks this page knows nothing, and an outage shorter than the gap can pass unrecorded. A day with no readings is drawn in the neutral and is never counted as a day that was up.

The last check landed Sep 6, 2026 - 00:51 UTC, 0 seconds ago, and checks have been arriving about every 2 seconds. That interval is measured from the readings rather than taken from the schedule that asks for them.