Episodes a provider genuinely has no data for stay actionable forever
and, via syncRefreshDebtForItem, pin the whole series item at
RefreshDebtReasonEpisodeIncomplete -> priority 300 (the top band). Those
debt rows never drain and always cut the queue ahead of legitimately-due
content, every backoff interval, indefinitely.
Classify episode-incomplete debt as terminal once its persistent
attempt_count reaches a small cap (3, ~5 days under the existing stepped
backoff): demote it off the priority-300 band (to a still-fixable
reason's band if one is present, else a terminal floor) and re-check it
rarely (90d) instead of every 30d. The signal is purely attempt_count on
the existing row -- no schema change. Demote rather than delete: deleting
would let the next series pass recreate the row at attempt_count 0 and
restart the cycle.
Giving up fast is safe: the library "refresh incomplete" and per-item
refresh endpoints re-fetch on demand and bypass the backoff, and
media_items.episode_metadata_incomplete is left untouched so they still
find these items. Logs the terminal transition (no silent demotion).
AI-use disclosure: implemented with AI assistance (Claude Code).
