Replay: fail fast on a dead AI account; extraction reports to ai_health

First replay (290 calls, 09-22 10:00-12:00 ET) produced 0 incidents and
no errors: every gpt-4o-mini extraction failed and _sync_extract
swallowed it as "no scenes". Same shape as #169 — and the live extraction
tier in /health/ai had no reporter at all, so this has been invisible in
production too.

- intelligence: API failures propagate out of _sync_extract; extract_scenes
  reports them to ai_health ("extraction" tier, billing/dead-model
  classified) and still returns [] so the pipeline degrades as before.
- ai_health: inside a replay sandbox, failures go to the run's own sink
  instead of being dropped.
- replay: aborts after 5 permanent failures on a tier, naming the cause;
  run metrics carry ai_failures; UI shows them.
- replay estimate: audio minutes from started_at/ended_at (no duration
  field exists on call docs).
- ReplayTab exposes the loaded run on window.__drbReplay for in-page
  analysis.

c2-core: 458 pass.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

drb-c2-core/tests/test_replay.py

@@ -369,3 +369,44 @@ async def test_replay_never_touches_live_ai_health_or_review_queue():
     finally:
         fstore.exit_sandbox(tok)
     assert ai_health.snapshot() == before
+
+
+@pytest.mark.asyncio
+async def test_run_aborts_when_an_ai_account_is_dead(store):
+    """An unfunded OpenAI account made the first smoke run a sandbox of 290
+    orphans that looked like a result. A permanently failing tier now stops
+    the run and names the cause."""
+    store.data["calls"] = {
+        f"call-{i}": _live_call(i, i, "Car 12 responding to an MVA on Main Street") for i in range(1, 30)
+    }
+
+    def broke(*a, **kw):
+        raise RuntimeError("Error code: 429 - You exceeded your current quota (insufficient_quota)")
+
+    with patch("app.internal.intelligence._sync_extract", broke), \
+         patch("app.internal.intelligence.classify_chatter", return_value=(False, None)):
+        run = await replay.start_run(
+            org_id="org-1", date_from=T0 - timedelta(hours=1), date_to=T0 + timedelta(hours=1),
+            mode="transcripts", system_ids=None, source_run_id=None, label="", actor="t")
+        await replay._active_task
+
+    run = store.data["replay_runs"][run["run_id"]]
+    assert run["status"] == "failed"
+    assert any("out of credit" in e for e in run["errors"])
+    assert run["progress"]["done"] < 29
+
+
+@pytest.mark.asyncio
+async def test_live_extraction_failure_reports_to_ai_health():
+    from app.internal import ai_health, intelligence
+
+    def broke(*a, **kw):
+        raise RuntimeError("insufficient_quota")
+
+    with patch.object(intelligence, "_sync_extract", broke), \
+         patch.object(ai_health, "report_degraded") as degraded, \
+         patch.object(fstore, "doc_set"), patch.object(fstore, "doc_get_cached", return_value=None):
+        scenes = await intelligence.extract_scenes("c1", "Car 12 responding to an MVA on Main Street")
+    assert scenes == []
+    assert degraded.call_args.args[0] == "extraction"
+    assert degraded.call_args.kwargs["permanent"] is True