Doc 04 — Branch E: Live Inventory vs Master Design Rev5 Reconciliation

Verdict: PASS. The live IU substrate is materially MORE ready than Master Design Rev5 assumed. Several design docs label as "paper" / "future" things that are now live and proven. Patches are documentation-truth patches, not blockers — current limited-pilot readiness does NOT depend on any of them.

Legend: Patch? = does Master Design need updating · Pilot dep? = does limited-pilot readiness depend on the patch.

#	Design assumption (Rev5 / earlier)	Live reality (2026-05-28)	Impact	Patch?	Pilot dep?
1	b/c/f filters are conceptual / harness TBD	dot_iu_filter_axis_b, dot_iu_sql_link_validate, dot_iu_sql_link_resolve, dot_iu_test_harness_run live (read, non-mutating); axis-C via fn_iu_subtree	filters are DOT-callable today	YES (mark live)	no
2	Bounded gate protocol = design-first proposal	fn_iu_gate_open/close/verify_closed/watchdog + iu_gate_transition table live; fail-closed; all_safe=true	gate protocol is operational infra	YES (mark live)	no
3	Durable governed run untested / future	First committed non-ROLLBACK governed run done; durability proven via fresh connection; lawful retire ×3	governed mutation lifecycle proven end-to-end	YES (mark proven)	no
4	review_decision = open TD-P1 / undefined	TEST builder fn_iu_test_review_decision_create + dot_iu_test_review_decision_create live; split/merge FK-probe it	test path live; production path still undefined (Branch C)	YES (record test/prod split)	partial (prod blocks full prod only)
5	emit_enabled / event gate = conceptual	piece_event_runtime.emit_enabled is a real governable gate inside all 4 gate fns; emit proven under protocol	emission is a first-class governed gate	YES (mark governable)	no
6	DLQ replay = future	fn_iu_route_dead_letter_replay proven {replayed,dry_run,resolved}; gated by master routes gate	DLQ replay operational	YES (mark proven)	no
7	P-pub = will be enforced at birth	P-pub1/P-pub2 warn-only in L1 (BEFORE INSERT); fn_iu_create writes no authority_ref; 133/73 rows missing	enforcement is staged, not done; Rev5 should not assume hard P-pub	YES (record warn-only + staged plan)	partial (blocks full prod only)
8	KG / Đ39 relationships = to be built	universal_edges (2,199 edges, full Đ39 columns) live but integer-keyed & excludes IU; iu_relation uuid-native (60 contains) lacks provenance/confidence	generic edge table exists; IU not in it; IU edges lack Đ39 enrichment	YES (record gap + projection design)	no
9	Master Design "iu_relation" lineage types populated	only contains (60) populated; derived_from/supersedes/merged_from/split_from defined-but-empty	lineage edges minted only on real cuts	minor (note)	no
10	vocab.publication_* assumed sufficient	only incomex_council authority + law/design_doc types exist	vocab too thin for P-pub block	YES (vocab expansion is a Đ32 item)	no

Net reconciliation

• Direction of drift: live is ahead of design on operational machinery (gates, DOT wrappers, durable run, DLQ, emit gate); behind design only on the two governance hardening items (production review_decision, P-pub hard enforcement) and on IU-KG enrichment.
• Master Design Rev5 patch set (documentation truth): items 1,2,3,5,6 → "live/proven"; item 4 → record test-vs-prod split; item 7 → warn-only + staged plan; item 8 → KG projection design + Đ39 enrichment gap; item 10 → vocab expansion as Đ32 work.
• Readiness dependency: none of these patches gate the limited pilot. They keep the design honest and feed the next macros. Items 4 and 7 gate full production (already authority-packed in docs 01/02).
• Recommended carrier: MASTER_DESIGN_LIVE_INVENTORY_RECONCILIATION_PATCH_*X (doc 06, macro 6) — a patch_document pass over Rev5, not a re-upload.