RS5B-CLOSEOUT-PATCH1 07 — Impact Map & Superseded Closeout Wording — 2026-06-21

Purpose: record every closeout phrase Codex rejected, the problem, the new controlling phrase, and status — so the consolidated reading is unambiguous about which text governs. Supersessions are wording/pointer changes carried additively; the original rev-1 closeout files are not mutated. Job A is not reopened. Gate: REGISTRATION_HOLD · REGISTRATION_CAN_PROCEED = NO · 0 mutations.

Columns: old phrase / source · problem · new controlling phrase · status.

---

1. Destructive rollback lines (Codex §5 — superseded by file 02 / file 03)

M1 — C1 "drop/disable vocabulary contract"

• Old phrase / source: closeout 06 §1 P2-C1 — "Rolled back separately: drop/disable the vocabulary contract without touching effect/digest carriers."
• Problem (Codex §5.1): dropping a canonical_operation value/contract referenced by C2 orphans or makes existing C2 identities uninterpretable.
• New controlling phrase: versioned retire/supersession with historical resolution — keep retired value resolvable; new use fail-closed unless successor mapping (file 02 C1).
• Status: SUPERSEDED_BY_RS5B_CLOSEOUT_PATCH1.

M2 — C3 "revoke/drop ownership row"

• Old phrase / source: closeout 06 §1 P2-C3 — "Rolled back separately: revoke/drop a single ownership row independently."
• Problem (Codex §5.2): a referenced owner/scope row cannot be dropped without resolving dependent bindings, audit history, prior decisions; revocation ≠ deletion.
• New controlling phrase: revoke/supersede preserving row identity + audit; future authority fail-closed if revoked (file 02 C3). The "revoke" intent survives; the "drop" branch is removed.
• Status: SUPERSEDED_BY_RS5B_CLOSEOUT_PATCH1.

M3 — C4 "drop hash record"

• Old phrase / source: closeout 06 §1 P2-C4 — "Rolled back separately: drop a hash record without touching the digest schema."
• Problem (Codex §5.3): C2/effect_identity/audit reference the hash; destructive removal breaks reproducibility and audit.
• New controlling phrase: preserve hash evidence; supersede only for new use (mark superseded/invalid-for-new-use) (file 02 C4).
• Status: SUPERSEDED_BY_RS5B_CLOSEOUT_PATCH1.

M4 — C5 "disable policy ref"

• Old phrase / source: closeout 06 §1 P2-C5 — "Rolled back separately: disable one policy reference without affecting the others."
• Problem (Codex §5.4): disabling a referenced policy without a compatibility/supersession rule leaves old C2 references unresolved or silently re-meant.
• New controlling phrase: policy version supersession + compatibility; preserve prior policy text/ref; new use requires successor (file 02 C5).
• Status: SUPERSEDED_BY_RS5B_CLOSEOUT_PATCH1.

M5 — C6 "disable the replay surface"

• Old phrase / source: closeout 06 §1 P2-C6 — "Rolled back separately: disable the replay surface without touching effect identity."
• Problem (extends Codex §5 reasoning): disabling/resetting the replay surface can make a consumed nonce reusable and lose replay-decision audit — an authority/safety regression.
• New controlling phrase: retire nonce policy version/issuer, preserve consumed-nonce + replay-decision audit; no nonce reuse becomes possible (file 02 C6).
• Status: SUPERSEDED_BY_RS5B_CLOSEOUT_PATCH1.

M6 — C7 "disable approval-as-a-check"

• Old phrase / source: closeout 06 §1 P2-C7 — "Rolled back separately: disable approval-as-a-check without affecting owner/registration carriers."
• Problem (Codex §5.5): disabling approval-as-check can retroactively weaken envelopes authorized under approval-required mode and erase approval evidence.
• New controlling phrase: versioned policy transition preserving prior approval-required envelopes + evidence; new mode explicit, forward-only (file 02 C7).
• Status: SUPERSEDED_BY_RS5B_CLOSEOUT_PATCH1.

2. Rollback-test wording (Codex §5 final ¶, §7 — superseded by file 04)

M7 — XBI-7 / R6 "rollback present"

• Old phrase / source: closeout 05 R6 ("per-carrier rollback plan … No carrier without a rollback surface") + closeout 07 XBI-7 ("rollback omitted ⇒ ROLLBACK_PLAN_ABSENT") — the only rollback test was presence.
• Problem (Codex §5 final ¶, §10.3): "they do not reject a destructive, dependency-unsafe, or history-erasing plan"; XBI-7 "checks only plan absence."
• New controlling phrase: rollback-validity oracle RBP-0..RBP-9 / RBP-PASS (file 04). ROLLBACK_PLAN_ABSENT survives as the single lowest-severity branch (RBP-1); destructive predicates RBP-2..RBP-9 outrank it and PASS. New fixtures XBI-11..XBI-19 cover orphan / history-erasure / authority-weakening / dependent-reference breakage.
• Status: SUPERSEDED / EXTENDED_BY_RS5B_CLOSEOUT_PATCH1 (R6 now requires rollback validity, not presence; XBI-7 retained only as RBP-1).

3. C7 conditionality (Codex §6 — superseded by file 05)

M8 — "all seven carriers must exist and PASS before P3"

• Old phrase / source: closeout 06 §4 — "All seven carriers above are pre-runtime prerequisites … must exist and PASS before P3," contradicting closeout 06 §1 P2-C7 "only if approval is used."
• Problem (Codex §6): C7 simultaneously optional and an unconditional prerequisite; "both cannot control simultaneously."
• New controlling phrase: deterministic approval_mode rule (file 05): C1–C6 unconditional pre-runtime prerequisites; C7 prerequisite iff approval_mode = APPROVAL_USED; APPROVAL_NOT_USED_BY_POLICY requires policy proof (APPROVAL_MODE_POLICY_UNPROVEN otherwise).
• Status: SUPERSEDED_BY_RS5B_CLOSEOUT_PATCH1. Any future "all seven … before P3" sentence lacking C7 conditionality is, by file 05 §4, superseded.

4. Gate sequencing (Codex §4 — superseded by file 06)

M9 — R12 dual-object review

• Old phrase / source: closeout 05 R12 — "an independent Codex/GPT review must accept the P2 entry-gate design and the specific build plan."
• Problem (Codex §4): mixes two reviews of two different objects in one requirement; the package has no carrier-specific build plan, so it cannot satisfy the plan-specific half.
• New controlling phrase: split — entry-gate-design review = Gate A input 2; specific-build-plan review = Gate B input 7 (file 06).
• Status: SUPERSEDED_BY_RS5B_CLOSEOUT_PATCH1.

M10 — output P2_ENTRY_DESIGN_READY_FOR_INDEPENDENT_REVIEW

• Old phrase / source: closeout 05 §2 all-pass output P2_ENTRY_DESIGN_READY_FOR_INDEPENDENT_REVIEW.
• Problem (Codex §4): circular — names "ready for independent review" although R12's independent review was already required to pass; and implies accepting baseline + a generic Chairman token could open P2.
• New controlling phrase: two distinct outputs — P2_BASELINE_ENTRY_DESIGN_ACCEPTED (Gate A; permits only preparing a plan) and P2_OPEN_AUTHORIZED_FOR_NAMED_CARRIER_PLAN_ONLY (Gate B; requires plan + rollback proof + preflight + scoped Chairman token + plan review) (file 06).
• Status: SUPERSEDED_BY_RS5B_CLOSEOUT_PATCH1.

5. Explicitly NOT superseded / NOT reopened

PATCH1 changes only the Job-B rollback/C7/gate wording above. The following remain exactly as accepted (Codex §3, §8, §9) and are not reopened:

• the consolidated RS5B_CONTRACT_AS_CORRECTED_BY_PATCH1_PATCH2 (closeout file 02) — all ten Job-A controlling points;
• supersessions S1–S4 (closeout file 03);
• pure effect_identity, authorization_binding_digest field set incl. effect_identity, canonical_operation, layered BI-E6/BI-E1 oracle, BINDING_CHECK_PASS necessary-not-sufficient;
• the 13-element authorization packet structure (item 13 Chairman final/unconditional);
• caveats C1–C5 (closeout file 04) — carried forward unchanged;
• closeout XBI-1..XBI-10 — retained; PATCH1 only adds XBI-11..XBI-25;
• RS4A-PATCH2-02 register_dot semantics, RS5A-PATCH4 quorum/G02/total-Q-order, candidate models, U1/U2/U3, owner/bootstrap/handler posture;
• the LEGO rule (born/test/change/rollback/join separate; no mega-registry) — PATCH1 strengthens rollback within it (invariant I9), does not relax it.

6. Net controlling reading

After PATCH1: Job A unchanged; Job B's rollback semantics are dependency-safe by invariant (file 02) over an explicit dependency graph (file 03), tested by a validity oracle that rejects destructive-but-present plans (file 04); C7 is conditional on approval_mode (file 05); and the gate is two clean stages (file 06). No original file overwritten; REGISTRATION_HOLD retained.