A switch ended by Cluster removal shows as the latest change on the next Cluster#4098

opened1d agobyNNick

GitHub writeback is off. Issues and comments you change here stay in GitCafe. Sync settings

latestChange (dashboard/src/modules/ingress/doors.server.ts) reads across the whole Organization, so an ingress_doors_attempt from a removed Cluster is still the latest change on the next one. Since #4037's fix, a switch open at removal ends failed: "The Cluster was removed."; the new Cluster's Network page shows it as a red alert, and canSwitchBack (WebTrafficSection.tsx) is true after a fresh setup, offering "Switch back" on a Cluster that never switched. Finished and failed attempts from before the removal already carry over the same way.

Likely fix: limit the latest change to attempts requested after the current pairing was created.

Scores (AGENTS.md bug rule): F 1 (needs a switch pending or running at removal, then a re-found Cluster), S 2 (cosmetic, one stray alert and button). F×S = 2, so filed, not fixed.