11000x · 05 — 8th healthcheck surface

What's added

cutter_agent/iu_core/healthcheck.py (the 5000x deploy-ready healthcheck) is bumped from 7 → 8 surfaces. The new 8th surface, piece_event_runtime, wraps the in-DB fn_iu_piece_event_runtime_healthcheck() so the same Python report covers it.

Surface	Source	Verdict rule
1. three_axis_cache	live drift view	unchanged
2. directus_collection	iu_three_axis_envelope + permissions	unchanged
3. qdrant_collection	iu_vector_sync_point indexed count	unchanged
4. auto_refresh_trigger	refresh_log + error_log	unchanged
5. vector_boundary	per-IU chunk_index NOT EXISTS GROUP BY HAVING	unchanged
6. write_gates	6 dot_config gate keys	unchanged
7. operator_runtime	dot_iu_command_run + dot_iu_runtime_lease	unchanged
8. piece_event_runtime	fn_iu_piece_event_runtime_healthcheck() flattened to a 9-column row	new

8th-surface verdict logic

def _verdict_piece_event_runtime(row):
    ok = bool(row.get("ok"))
    if not ok:
        return False, "piece_event_runtime substrate integrity failure"
    return True, (
        f"emit_enabled={row['emit_enabled']} "
        f"dry_run_only={row['dry_run_only']} "
        f"types={row['event_types']}"
    )

The verdict treats ok=true (substrate intact) as healthy regardless of gate state — the gate being false is the desired production state; flipping it open is an operator-driven phase, not an unhealthy condition.

If the substrate is broken (e.g. someone manually drops the trigger), ok=false is returned and the entire healthcheck exits with code 2 (FAIL).

In-DB verdict shape

{
  "surface": "piece_event_runtime",
  "ok": true,
  "view_exists": true,
  "fn_emit_exists": true,
  "fn_trg_exists": true,
  "trigger_exists": true,
  "gate_keys": 2,
  "piece_event_types_active": 6,
  "emit_enabled": "false",
  "dry_run_only": "true"
}

Failure modes detected:

• view_exists=false (someone DROPped v_piece_event_outbox)
• fn_emit_exists=false (someone DROPped fn_iu_piece_emit_event)
• fn_trg_exists=false (someone DROPped fn_iu_lifecycle_log_emit_piece_event_trg)
• trigger_exists=false (someone DROPped the trigger)
• gate_keys<>2 (someone DELETEd a piece_event_runtime.* config row)
• piece_event_types_active<>6 (someone retired one of the 6 piece.* events)

Any one of these returns ok=false.

Tests updated for the 8th surface

Test	Bumped from → to
test_iu_core_5000x_healthcheck.py::TestSurfaceCoverage::test_expected_surface_count	7 → 8
test_iu_core_5000x_healthcheck.py::TestHappyPath::test_all_surfaces_ok	uses HAPPY_RESULTS including new piece_event_runtime happy row
test_iu_core_5000x_healthcheck.py::TestHappyPath::test_json_output_shape	decoded surfaces count 7 → 8
tests/test_iu_core_piece_event_runtime.py::TestHealthcheckEighthSurface	2 new tests on the verdict function + module docstring

All passing.

Mac cron pilot impact

The 11000x healthcheck still uses the same iu_core_healthcheck_wrapper.sh from ops/healthcheck-cron-package/scripts/. The wrapper invokes python3 -m cutter_agent.iu_core.healthcheck which now reports 8 surfaces. No script changes needed.