A shared error-reporting layer plus post-publish smoke tests that auto-revert a release when checks fail.
Before this, every app reported errors its own way, and a bad release was caught by whoever happened to be watching. I standardized how critical frontend errors get reported and added post-publish checks that revert a release on their own when it regresses.
The pieces are small and shared: a required error boundary, a network helper, and one way to report a critical error — all feeding the same telemetry the post-publish checks read. So a regression shows up in the signal a release is judged on, not in a bug report three days later. This lives in the Frontend Platform stack the whole frontend org builds on.
Highlights
- Required error boundary that captures unhandled exceptions as critical signals
- Network helper that reports failed requests automatically
- One shared helper for reporting critical errors across apps
- Post-publish workflow that runs smoke tests and reverts on failure
Result
Promotions gate on smoke-test results instead of a manual sign-off, and a release that fails them reverts automatically — so the window where a bad build is live shrinks to the length of the check, not the length of someone noticing.