The Cloudflare Worker That Ran Perfectly and Still Failed Twice

A Cloudflare Worker that monitors other Workers uncovered two reliability bugs during review.

One fail-open bug could let a KV read failure abort an entire hourly check instead of just disabling maintenance-window suppression.

After adding Sentry and cron monitoring, the team found the monitor still wasn’t showing up in Sentry, meaning the health check was broken.

The case shows how code can deploy cleanly yet still miss its real job without explicit error handling and observability.