02 — F-1 / F-2 Consolidated Findings

A single ledger of what the staircase has proven so far, so F-3 builds on facts rather than re-deriving them. Each row notes whether F-3 re-confirmed, extended, or carried forward the finding.

2.1 Proven facts (cumulative)

#	Finding	Established	F-3 disposition
P1	Coverage path inventory → effective-owner → gap is sound and honest; empty ownership ⇒ maximal honest gap set, zero false PASS	F-1 (F-DR-2)	re-confirmed at 10-subset and full-35
P2	Gap grain = (object × active-scope); ceilings are N×6	F-1 (F-DR-1)	extended: 60 (10×6) and 210 (35×6) proven live
P3	Inventory seam wireable now at collection grain (decision-gated, not engineering-gated)	F-1 (F-DR-3)	re-confirmed; full-35 form compiles live
P4	Seam wiring can be real view objects inside the txn, auto-restored inert on ROLLBACK	F-2 (§2.3)	re-confirmed for both seams + independent pg_get_viewdef check
P5	F2-1 cast collection_name::text AS object_ref is load-bearing for any CREATE OR REPLACE	F-2 (F2-1)	used verbatim; both seams compiled
P6	Recursive resolution: direct (depth 0) + inherited (via is_inherited_anchor); gap honors inheritance	F-2 (F2-3)	extended: 1 group anchor → 4 inherited children; 2 direct; gap 60→55
P7	One-accountable invariant enforced by uq_gov_obj_accountable; conflict view=0 on real store, detectable by predicate on index-less data	F-2 (F2-4)	re-confirmed (E6=0, E7=1, N1 rejects)
P8	Integrity rails fail-closed: UNIQUE + 2×FK + CHECK(owner_kind) + CHECK(delegated TTL) + candidate ruleset FK	F-2 (F2-4)	extended to 7 (added lifecycle_status CHECK)
P9	Zero emit / zero issue / zero activation boundary holds	F-1/F-2	re-confirmed (gov_emit 0, sys_issues 198,442, no activation)
P10	Entry==exit, Δ=0, no residue, idle-tx=0	F-1/F-2	re-confirmed on two channels

2.2 Boundaries that remain (carried forward unchanged)

#	Boundary	Status after F-3
B1	Object-grain (birth_registry ~1.08M)	NOT attempted; needs SB-13 cursor; collection grain only — NO-GO for first commit
B2	Persistence / emit / activation (scanner, worker, event, DOT, handler, UI)	NOT attempted; F-5..F-7, ratification-gated
B3	Axis / assignment / topic / UI-exposed coverage	axis_registry/axis_assignment/coverage_rule ABSENT ⇒ fake/test-only; NO-GO
B4	The COMMIT itself (seam wiring crosses inert→active for coverage)	F-4; requires L2+L4 ratification; os_proposal_approvals=0 gate unsolved
B5	Performance at scale	10-subset + 210-row full inventory sub-second; says nothing about object grain

2.3 Documentation items reconciled

• F2-6 / F-DR-6 (missing GPT-review artifacts). The two review files named in the F-3 mission — gpt-review-f1-readonly-dry-run-pass-go-f2-rollback-only-2026-06-02.md and gpt-review-f2-rollback-only-pass-go-f3-consolidation-rollback-only-2026-06-02.md — are not present on disk (knowledge/dev/reports/architecture/) nor in the KB architecture path. Same pattern as F-1/F-2. No authorization gap: each GO is grounded in the published predecessor package's own 07/08 GO doc (F-1 doc 07 = GO F-2; F-2 doc 08 = GO F-3). Logged for operator reconciliation; the parallel GPT review may be pending or filed elsewhere.
• F2-7 (governing law file). knowledge/dev/laws/prompt-muc-tieu-mo-for-claude-code.md is absent/empty (the laws/ dir is empty on disk). Immaterial to F-3 — the governing principles (honest reporting, no hardcode, fail-closed, no hidden island, ratification-gated activation) are internalized and honored throughout. Flagged for the operator to restore canonical law text.

2.4 What F-3 newly adds (preview; detail in docs 04/05/08)

1. The 210 full ceiling proven live as a real view, not an estimate (closes the inference in F2-2).
2. Real acyclic containment from collection_registry."group" — first time inheritance is exercised over a genuine production source rather than a 1-edge fixture.
3. Group-level inheritance fan-out: one anchor on a folder-group resolves owners for all its child collections — the F-4-relevant scaling behavior.
4. A concrete F-4 seam-commit packet (doc 08): exact DDL, ceilings, forbidden set, stop conditions, rollback plan, ratification gate.