All systems operational.

The last few things that went wrong.

• 27 Jul 2026 Hourly data checks silently failing for almost four weeks

  A code change we shipped after NHSBSA rate-limited us on 26 July referenced a database column that did not exist, so every hourly check from 27 July crashed before doing any work. The crash landed in a spot none of our monitoring watched, which is the part we are not proud of. Fixed 22 August; the pipeline caught up the same day. No published figure was ever wrong, but the June data release reached the site later than our 48-hour promise. We now watch the heartbeat itself, and this page reads its dates from the pipeline rather than from a human remembering to update them.