fix(worker): reconcile every job + backfill owned releases

Wanted releases pressed while the monitor is off never got reconciled, so
they lingered on the wanted list. Extract finish_job() and call it for every
finished job (not just when monitor.enabled) so the MonitoredRelease
transitions to fulfilled. Add fulfill_owned_releases(), run once at startup,
to backfill releases already owned at/above the quality cutoff (self-heals
items stuck from before this fix). Both idempotent + tested.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

worker/lyra_worker/main.py

@@ -7,7 +7,7 @@ from lyra_worker.config import get_config
 from lyra_worker.db import wait_for_db
 from lyra_worker.discovery import DiscoveryConfig, run_discovery
 from lyra_worker.library import clear_staging_root
-from lyra_worker.monitor import MonitorConfig, reconcile, sweep
+from lyra_worker.monitor import MonitorConfig, fulfill_owned_releases, reconcile, sweep
 from lyra_worker.pipeline import run_pipeline
 from lyra_worker.registry import (
     build_adapters, build_browser, build_probe, build_resolver,
@@ -98,6 +98,17 @@ def _run_discovery(conn, sources, browser, config=None) -> None:
     print(f"worker: discovery done — {result.artists} artists, {result.albums} albums", flush=True)
 
 
+def finish_job(conn, job_id, mcfg) -> None:
+    """Feed a finished job's outcome back into its MonitoredRelease. Runs for EVERY job,
+    monitor enabled or not: a manually-wanted release must still transition to fulfilled so
+    it drops off the Wanted list. A no-op for plain (unlinked) requests."""
+    try:
+        reconcile(conn, job_id, mcfg)
+    except Exception as e:  # reconcile must never take down the worker loop
+        print(f"worker: reconcile failed for job {job_id}: {e}", flush=True)
+        conn.rollback()
+
+
 def maybe_run_discovery(conn, sources, browser, config, dcfg, now, last_discover_tick,
                         tick_seconds: float = DISCOVER_TICK_SECONDS) -> float:
     """Run discovery at most once per loop iteration, whether triggered by the schedule
@@ -127,6 +138,11 @@ def run_forever() -> None:
     browser = build_browser()
     probe = build_probe()
     similarity_sources = build_similarity_sources(get_config(conn))
+    # Backfill: releases pressed while the monitor was off never got reconciled, so they linger
+    # on the wanted list. Fix any already-owned ones once at startup.
+    fixed = fulfill_owned_releases(conn, MonitorConfig.from_config(get_config(conn)))
+    if fixed:
+        print(f"worker: reconciled {fixed} already-owned release(s) to fulfilled", flush=True)
     print(f"worker: {len(adapters)} adapter(s) enabled: {[a.name for a in adapters]}", flush=True)
     print("worker: waiting for jobs", flush=True)
     last_tick = 0.0
@@ -167,12 +183,7 @@ def run_forever() -> None:
                 except Exception as e:  # one bad job must not take down the worker loop
                     print(f"worker: pipeline failed for job {job_id}: {e}", flush=True)
                     conn.rollback()
-                if mcfg.enabled:
-                    try:
-                        reconcile(conn, job_id, mcfg)
-                    except Exception as e:
-                        print(f"worker: reconcile failed for job {job_id}: {e}", flush=True)
-                        conn.rollback()
+                finish_job(conn, job_id, mcfg)
                 print(f"worker: finished job {job_id}", flush=True)
             except psycopg.OperationalError as e:
                 print(f"worker: db connection lost ({e}); reconnecting...", flush=True)