VERIFY — Registries-Pivot Process Coverage Closeout · Independent Live-Evidence Recheck

Date: 2026-06-04 · Mode: VERIFY_ONLY (read-only; no production mutation) · Macro: VERIFY_ONLY_RP_PROCESS_CLOSEOUT_LIVE_EVIDENCE_RECHECK Subject checkpoint: checkpoint-rp-process-coverage-ui-owner-canon-closeout-2026-06-04 Method: independent live reads via query_pg (READ ONLY role, statement_timeout 5s) against directus DB. No writes, no DDL, no fn calls.

FINAL STATUS: PASS_CONFIRMED

Every substantive engineering count and safety invariant in the checkpoint reproduces against the live DB. The single numeric divergence — birth_registry +20 vs the recorded baseline — is fully attributable to non-engineering KB-document births and does not contradict the checkpoint's stated invariant ("birth-free DB-eng"). See reconciliation below.

---

1. The 3 birth-free views exist

pg_views confirms all three present:

• v_workflow_residual_evidence_hardening_v3 ✓
• v_process_axis_owner_decision_flow ✓
• v_rp_process_canon_gate_summary ✓

2. Live counts vs checkpoint

Metric	Checkpoint	Live (2026-06-04)	Verdict
birth_registry	1,168,759	1,168,779 (+20)	RECONCILED — all +20 are KB-doc births (see §4)
trigger_guard_alerts	129	129	MATCH ✓
approval_requests	230	230	MATCH ✓
axis_assignment (total)	25	25	MATCH ✓
official AX-PROCESS RP assigned	0 / 453	0 / 453	MATCH ✓
candidate-visible	69	69	MATCH ✓
owner decision flow rows	12 (5 president + 7 owner)	12 (5 PROC-OWN president-vote + 6 WPC owner + 1 docker-runtime owner)	MATCH ✓
residual actionable	34 → 19	19 (11 docker owner-gated + 8 owner-reconcile)	MATCH ✓
residual evidence-gated	23 → 0	0 (ledger v3 = 23 dispositioned rows: 9 already-dot + 4 db-poller + 2 not-process = 15 resolved, + 8 owner-reconcile)	MATCH ✓
job:cut	VERIFIED, sole, 8 members	verified_candidates = 1 (job:cut, 8 members)	MATCH ✓
dot:kg	correlated_dryrun_observed	2 DRY_RUN observations (process_run_observation)	MATCH ✓
REAL_RUN	0	0 (only 2 DRY_RUN + 4 SIMULATED_DRY_RUN)	MATCH ✓
event_type_registry process.* active	0 / 7	0 / 7 (all active=false)	MATCH ✓
governance_object_ownership	0	0	MATCH ✓
process_axis_delegation	0	0	MATCH ✓
canon gate verdict	CANON_BLOCKED_OWNER_ONLY	CANON_BLOCKED_OWNER_ONLY	MATCH ✓

3. Independent base-table confirmations (not just summary views)

• axis_assignment — all 25 rows are AX-TOPIC (TOPIC-CAND:*), status=candidate, approval_ref=NULL, approved_by=NULL, source MACRO:topic-axis-ratify-apply-pilot. Zero AX-PROCESS rows. → official AX-PROCESS RP = 0/453 is true at the base table; not faked by a view.
• v_workflow_residual_evidence_hardening_v3 (23 rows) dispositions tally exactly: ALREADY_DOT_MANAGED=9 (maps_to_dot=true), COMPONENT_DB_POLLER=4, NOT_PROCESS (infra-config + template)=2, OWNER_RECONCILE_REGISTER_OR_QUARANTINE=8. 15 resolved + 8 awaiting owner reconcile = 23. (Actionable residual 19 = 8 reconcile + 11 docker-runtime owner-gated.)
• process_run_observation grouped by evidence_type: DRY_RUN=2, SIMULATED_DRY_RUN=4, REAL_RUN=0.
• event_type_registry process.*: 7 rows (run.started/completed/failed, step.started/completed, correlation.missing, evidence.missing) — all active=false.
• wf_process_candidate (19 WPC rows): every row lifecycle_status=DRAFT, owner_status=UNASSIGNED, dryrun_realrun_evidence=NONE (0 DRY_RUN/0 REAL_RUN). No candidate confirmed or birthed.
• governance_object_ownership=0, **process_axis_delegation=0`.

4. birth_registry +20 reconciliation (why this is NOT a mismatch)

The most-recent 25 birth_registry rows are all collection_name=knowledge_documents, species_code=ai_support, dot_origin=PG:trg_birth_knowledge_documents (entity_code knowledge_documents::6571 … ::6595), born 2026-06-04 10:49–11:12 UTC. These are KB-document births triggered by content/report uploads — exactly the documented "+20 KB-doc births" side effect ("KB uploads cause births ≠ DB-eng").

No engineering births occurred: there are no new births in axis_assignment, dot_tools, pivot_definitions, governance_object_ownership, or apr_approvals collections in this window. The checkpoint's invariant is stated precisely as "birth-free (DB-eng)" — that invariant holds. The global birth_registry counter is expected to keep ticking with documentation births (including the report+checkpoint produced by this verification, which add 2 more KB-doc births).

5. Safety invariants — all confirmed PASS

• ✅ No new birth/canon outside documented KB docs (the only births are knowledge_documents).
• ✅ No approval approved — approval_requests=230 unchanged; no AX-PROCESS approval_ref set.
• ✅ No event activation — process.* 0/7 active.
• ✅ No source-IU edit — axis_assignment all AX-TOPIC candidates from prior pilot, unchanged; no IU writes.
• ✅ No REAL_RUN — process_run_observation REAL_RUN=0.
• ✅ No fake official RP — AX-PROCESS assignment 0 at the base table.
• ✅ No fake ownership / no fake delegation — both tables empty (0).

6. Note on the residual view vs the "19" figure

v_workflow_residual_evidence_hardening_v3 returns 23 rows (the evidence-hardening ledger of the 23 previously evidence-gated items), which is a different metric from the "19 actionable residual" (8 owner-reconcile + 11 docker-runtime). Both are internally consistent and both match the checkpoint; the 23 is the disposition ledger, the 19 is the still-actionable subset. Flagged here to prevent a false "23≠19" reading.

Determination

PASS_CONFIRMED. Live DB reproduces the checkpoint on all 16 substantive metrics and all 7 safety invariants. The lone numeric delta (birth_registry +20) is reconciled to KB-document births and is consistent with the checkpoint's "birth-free (DB-eng)" claim. No production state was mutated by this verification (read-only query_pg); the only writes are this report and its checkpoint to the KB.