04 — Master Design Rev5 ↔ Live Inventory Reconciliation

The live substrate is ahead of the design's paper/gap assumptions. This is an actionable patch list: what the design says, what live says, the action, and the exact target doc to patch later. Patches are doc-truth only — no code/schema change here. Target design dir: knowledge/dev/design/v0.6-iu-4mothers-event-foundation-rev2/ (docs 00–10).

---

1. Reconciliation table

#	Design assumption (Rev5)	Live reality (2026-05-28)	Action	Target doc
R1	b/c/f tests need a harness to be built	5 harness fns live + 5 catalog rows (fn_iu_filter_axis_b/_c_subtree/_sql_link_validate_all/_resolve_all/_bcf_harness_run); catalog=52, run=55	update doc — mark harness EXISTS	00, 06
R2	Bounded gate protocol = design-first/not deployed	iu_gate_transition table + fn_iu_gate_open/close/verify_closed/watchdog LIVE; gates all closed, all_safe=true	update doc — Điều 5 Tầng3 state machine partially deployed (gate protocol now live)	00, 02, 09
R3	Durable governed run untested	First committed durable governed run proven (retire×3, durability via fresh connection)	update doc — durable run PROVEN	00, 06
R4	DLQ replay untested	fn_iu_route_dead_letter_replay proven (forced DLQ → replay); dlq=0 now	update doc — DLQ replay PROVEN; replay gated by routes-master, not dlq.replay_enabled	03
R5	No review_decision builder	fn_iu_test_review_decision_create LIVE (test_scope); production builder still absent	update doc — distinguish test vs prod builder (see doc 06)	04
R6	emit_enabled not integrated into gates	piece_event_runtime.emit_enabled is a governable gate in fn_iu_gate_verify_closed	update doc — emit gate integrated	03
R7	KG/relationships are paper	universal_edges (2199, integer-keyed, has provenance/confidence/valid_time) + iu_relation (60, UUID, containment) both LIVE	update doc — KG substrate exists; reconcile two-graph reality (doc 03)	04, 00
R8	P-pub authority enforced	P-pub1/P-pub2 are warn-only; fn_iu_create has no authority_ref param; 133/219 IUs lack authority publication	update doc — P-pub is warn-only, staged promotion needed (doc 05)	04, 06, 10
R9	Candidate Registry tables planned	6 CR tables (field/input_form/output_table/dot_function/task_template/tier_registry) ABSENT; many adjacent registries exist to extend	update doc — CR is paper; reuse map (doc 02)	04, 06
R10	4 Mothers register as Đ37 factories	governance_registry=5, all council/system, no factory rows, no output_target, no human roles	update doc — factory rows not yet created; human-org-role law missing (doc 07)	04, 10
R11	birth_registry modest	876047 rows — industrial-scale birth ledger already live	no action — confirms Đ0-G industrial-birth assumption	10
R12	species_collection_map planned	LIVE	update doc — exists	10
R13	gateway allowlist covers lifecycle writers	allowlist now includes fn_iu_retire,fn_iu_supersede (the prior-bundle fix persisted)	no action — already correct	00

2. Risk if stale design remains

• R7/R8/R9 are the dangerous ones. If the design still says "KG is paper" / "P-pub enforced" / "CR planned", a future implementer may (a) build a second IU graph (hidden SoT), (b) assume P-pub already blocks unauthorized births (it only warns), or (c) build CR tables that duplicate collection_registry/dot_iu_command_catalog. Each is a real correctness hazard.
• R2/R3/R4/R6 are lower risk (under-claiming maturity) but cause wasted re-implementation of already-live machinery.
• R10 risk: assuming factories are registered when they are not could let an unregistered Mother attempt to birth objects without Đ37 authority.

3. Patch sequencing (later macro, doc-only)

1. Patch 00-master-design-rev2.md + 07-master-design-rev2-report.md with the live-maturity deltas (R1–R6, R13).
2. Patch 04-iu-centered-4mothers-binding-design.md with R7–R10 (KG two-graph, P-pub warn-only, CR reuse map, factory-not-registered).
3. Patch 10-industrial-birth-cross-law-addendum.md with R10–R12.
4. Patch 03-event-5layer-realtime-dlq-design.md with R4/R6.
5. Use patch_document (exact old_str→new_str replay) on the KB copies — serialize same-doc patches; revision +1 per patch (treat the Status-line label as authoritative, not the integer).

4. What NOT to patch

• Do not "upgrade" the design to claim production-readiness it lacks (P-pub block_all, production review_decision, CR built, factories registered). Those remain future work; the design should describe them as authority-packed / deferred, matching docs 05/06/02/07.