fix(release-health): per-cadence report sinks, accept default catch-up window

The probe expected the daily fep_feedback_intake sink on all three
frontend-patterns definitions and required an explicit
catchup_window_seconds that live rows do not carry. It now checks each
cadence's own sink, and accepts an absent window, which resolves to the
catchup_latest 24h default (the schedule's catchup_window is still
checked). Closes ACTIVITY-WP-ADHOC-2026-10-19-T03.

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

Assistant: claude-code
Assistant-Model: opus
Assistant-Process: 244572@bnt-lap001
Assistant-Session: 3b235e1a-ae45-4d0f-ae70-17a9bd78ece8

workplans/ADHOC-2026-10-19.md

@@ -57,7 +57,7 @@ f9f2640 / `sha256:fdc631d7`. Evidence:
 
 ```task
 id: ACTIVITY-WP-ADHOC-2026-10-19-T03
-status: todo
+status: done
 priority: medium
 state_hub_task_id: "0bfc9af8-1b75-59ab-91e6-cef5b09bc08d"
 ```
@@ -71,3 +71,17 @@ three frontend-patterns definitions, while weekly and monthly use
 `fep_improvement_review` and `fep_monthly_review`. Fix the probe to check the
 per-cadence sinks, and decide whether to run an admin definition sync or
 accept a missing `catchup_window_seconds` when the schedule enforces it.
+
+2026-10-20: fixed in `release_health.py`.
+- Each cadence now expects its own sink: daily `fep_feedback_intake`, weekly
+  `fep_improvement_review`, monthly `fep_monthly_review`. All sit on topic
+  `fe2aaa78` with author activity-core, as in the 2026-10-03
+  natural-cadence evidence.
+- A row with no `catchup_window_seconds` is accepted, because the
+  `catchup_latest` policy's default in `schedule_manager` is 24h. An explicit
+  value other than 86400 is still refused, and the probe still checks the live
+  Temporal `catchup_window` separately. No admin definition sync is needed.
+- Two new tests cover these cases. The full suite passes (877 passed,
+  1 skipped).
+- Not deployed: the fix ships with the next image release, and live
+  `/release/health` stays false until then.