Reject runner updates for tasks that already left the pool

When a remote task finishes it is removed from TaskPool, so this new HydrateTaskRunnerFromDB fallback also rehydrates already-finished tasks from the database. In that case TaskRunner.SetStatus in services/tasks/TaskRunner_logging.go returns immediately for success/fail/stopped, but this handler still executes LogWithTime and SetCommit, so any delayed or retried runner PUT can append output or overwrite commit metadata on a completed task. This shows up in HA/load-balanced setups where the last progress request lands after the owning node has already stopped tracking the task.

Useful? React with 👍 / 👎.

@codex fix it

Summary

• Updated RunnerController.UpdateRunner to ignore runner progress payloads once the matched task is already in a finished state, preventing delayed log writes or commit metadata updates from mutating completed tasks. api/runners/runners.goL275-L299
• Added a regression test that exercises a late runner update against a finished task and verifies the handler returns 204 No Content without changing the stored commit metadata. api/runners/runners_test.goL27-L87
• Committed the follow-up fix on the current branch as 4cce352 and recorded a PR title/body for the new follow-up PR. api/runners/runners.goL280-L287 api/runners/runners_test.goL27-L87

Testing

• ✅ gofmt -w api/runners/runners.go api/runners/runners_test.go
• ✅ git commit -m "fix(runners): ignore late updates for finished tasks"
• ✅ go test ./pkg/task_logger -run TestDoesNotExist -count=1
• ⚠️ timeout 30s go test ./api/runners -run '^$' -count=1 -timeout 20s; echo EXIT:$? — timed out in this environment with EXIT:124.
• ⚠️ timeout 30s go test ./api/runners -run TestUpdateRunnerIgnoresFinishedTaskUpdates -count=1; echo EXIT:$? — timed out in this environment with EXIT:124.

View task →