📖About

Today on INI-029 attempt 18: Argus's tester crashed (real bug — Flask decorators on http.server app). Albus's troubleshooter started, took 169s to diagnose + DIRECT_FIX the bug. But Elrond's Dispatcher only checks for live recovery via executions in planning/executing status — not Albus's running droid_run. So once the 120s grace expired (and Albus was 49s into his work but mid-run), Dispatcher escalated INI-029 to blocked. Albus's fix landed but too late. Fix: extend _has_recovery_in_flight to ALSO return True when ANY al-auto-01 droid_run is status='running' on any exec for the initiative. Albus's 10-min self-timeout + 30-min orphan watchdog already bound how long this can wait, so no separate cap needed in the dispatcher.