history_coverage is only written at the end of an ingest, but
has_completion_data was reading from it. During the first scan the coverage
table is empty while watch_event already holds tens of thousands of Tautulli
rows, so the dashboard announced a fallback that had not happened and claimed
the rejection component was disabled when it was not.
The flag now comes from the events themselves — does any row carry a
percent_complete — which is true the moment Tautulli rows land and false for
Plex-only history. history_source falls back to the scan record when coverage
is absent, so it reads "tautulli" mid-scan instead of null.
Also: the banner named Plex as the source without checking, and now reports
whichever source is actually active, and stays quiet while a scan is running
since the counts are still moving.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GVbG48GAXfCZatcmX123Ra
Jess Hallsworth
71cfa31f09