RS5B-CLOSEOUT-PATCH2 04 — Single-Input/Single-Code XBI Corrections & New Fixtures — 2026-06-21

Scope: close Codex blockers B3 (XBI-13 split), B4 (XBI-14 split), B5 (XBI-19 split), B6 (add successor-but-not-fail-closed fixture), B7 (add C2-authority-weakening fixture). Every fixture is one input · one expected code · one invariant tested · one reason no other code competes · PASS/seal/digest possible? No · result REJECTED. No fixture lists two outcomes. Gate: REGISTRATION_HOLD · REGISTRATION_CAN_PROCEED = NO · 0 mutations. Design fixtures (DEFINED_NOT_EXECUTED), not executed tests. RUNTIME_MUTATION_REJECTED (RBP-0) short-circuits any fixture if a runtime write appears. Supersession: PATCH1-04 XBI-13 (single, → RBP-4), XBI-14 (multi-outcome), XBI-19 (multi-outcome) are SUPERSEDED_BY_RS5B_CLOSEOUT_PATCH2 by the split fixtures below (file 08 M13/M14/M15). PATCH1 XBI-11/12/15/16/17/18/20–25 are retained unchanged.

1. XBI-13 split (closes B3) — hash deletion vs hash-payload erasure

The single PATCH1 XBI-13 ("drop a hash referenced by effect_identity/audit ⇒ ROLLBACK_ERASES_HISTORY") contradicted the precedence (deletion of a referenced identity is RBP-2, which precedes RBP-4). It is split into two single-code fixtures discriminated by does the hash identity still resolve? (file 03 §2).

XBI-13A — hash identity deleted

Input (one): C4 rollback deletes the hash record/identity that an existing effect_identity and audit reference; the identity no longer resolves.
Expected code (one): ROLLBACK_DELETES_REFERENCED_IDENTITY (RBP-2).
Invariant tested (one): I1 stable identity (edge E3).
Why no other code competes: the identity itself is gone (file 03 RBP-2 card); orphan (RBP-3) and erasure (RBP-4) are downstream consequences but RBP-2 is rank 2 and dominates. RBP-2 precedes RBP-3/RBP-4.
PASS/seal/digest possible? No.
Result: REJECTED.

XBI-13B — hash identity remains, evidence payload erased

Input (one): C4 rollback keeps the hash record/identity resolvable (and its edge E3 resolves) but erases the stored hash payload / reproducibility data, so the historical artifact proof is unreadable.
Expected code (one): ROLLBACK_ERASES_HISTORY (RBP-4).
Invariant tested (one): I2 historical evidence / reproducibility (edge E3).
Why no other code competes: identity resolves (not RBP-2) and the edge resolves (not RBP-3); the meaning is unchanged, only the payload is unreadable (not RBP-5). Only RBP-4 names unreadable evidence while identity+edge survive.
PASS/seal/digest possible? No.
Result: REJECTED.

Distinct-code proof: XBI-13A → RBP-2 and XBI-13B → RBP-4 because they differ on the single discriminator "does the identity resolve?" (No vs Yes). They are disjoint inputs with disjoint codes. (B3 closed; self-review A11, file 09.)

2. XBI-14 split (closes B4) — orphaned reference vs re-meant reference

The PATCH1 XBI-14 listed both ROLLBACK_ORPHANS_DEPENDENCY and ROLLBACK_CHANGES_HISTORICAL_SEMANTICS. Split by does the old C2 reference still resolve?

XBI-14A — policy ref disabled, old C2 reference cannot resolve

Input (one): C5 rollback disables a referenced status/U3/audit policy so an old C2 *_policy_ref cannot resolve (the reference dangles); the policy object is not deleted as an identity.
Expected code (one): ROLLBACK_ORPHANS_DEPENDENCY (RBP-3).
Invariant tested (one): I3 reference integrity (edge E4).
Why no other code competes: no identity is deleted (not RBP-2); the reference does not resolve at all, so there is no surviving meaning to change (not RBP-5); no payload is selectively erased while the edge resolves (not RBP-4). Only RBP-3 names a dangling edge.
PASS/seal/digest possible? No.
Result: REJECTED.

XBI-14B — old C2 reference resolves but policy meaning changed

Input (one): C5 rollback keeps the old C2 *_policy_ref resolvable, but the slot now points at a re-meant policy, so the old envelope silently means something different.
Expected code (one): ROLLBACK_CHANGES_HISTORICAL_SEMANTICS (RBP-5).
Invariant tested (one): I4 semantic immutability (edges E4/E7).
Why no other code competes: the reference resolves (not RBP-3) to a readable policy (not RBP-4) and no identity is deleted (not RBP-2); the fault is reinterpretation. Only RBP-5 names silent meaning change.
PASS/seal/digest possible? No.
Result: REJECTED.

Distinct-code proof: XBI-14A → RBP-3 (does not resolve) and XBI-14B → RBP-5 (resolves but re-meant) differ on "does the reference resolve?" RBP-3 (rank 3) precedes RBP-5 (rank 5), consistent with the inputs. (B4 closed; self-review A12, file 09.)

3. XBI-19 split (closes B5) — non-local rewrite vs meaning-changing rewrite

The PATCH1 XBI-19 described ROLLBACK_NOT_LOCAL but acknowledged RBP-5 when meaning changes. Split by does the silent cross-carrier rewrite change prior meaning?

XBI-19A — silent C2 reference rewrite, meaning preserved

Input (one): a carrier rollback only "works" if C2 digest/envelope references for already-emitted effects are silently rewritten, but the rewrite does not change the prior envelope meaning (e.g. a pointer is re-homed to an identical-meaning record).
Expected code (one): ROLLBACK_NOT_LOCAL (RBP-10).
Invariant tested (one): I9 locality (the mega-registry tell).
Why no other code competes: the producer's own surface is fine and prior meaning is unchanged (not RBP-5); no identity deleted (not RBP-2), no orphan (not RBP-3), no erasure (not RBP-4). The sole fault is the cross-carrier silent mutation requirement — only RBP-10 names non-locality.
PASS/seal/digest possible? No.
Result: REJECTED.

XBI-19B — silent C2 rewrite that changes prior envelope meaning

Input (one): the carrier rollback silently rewrites C2 and the rewrite changes the meaning of a prior envelope.
Expected code (one): ROLLBACK_CHANGES_HISTORICAL_SEMANTICS (RBP-5).
Invariant tested (one): I4 semantic immutability.
Why no other code competes: although the input is also non-local (RBP-10), RBP-5 (rank 5) precedes RBP-10 (rank 10); the dominant, earliest-matching harm is the silent meaning change. The fixture emits the single earliest code per the precedence, not two.
PASS/seal/digest possible? No.
Result: REJECTED.

Distinct-code proof: XBI-19A → RBP-10 (meaning preserved) and XBI-19B → RBP-5 (meaning changed) differ on "does the rewrite change prior meaning?" Each emits exactly one code by precedence. (B5 closed; self-review A13, file 09.)

4. New fixture XBI-26 (closes B6) — successor exists but retired value still admissible

Input (one): a rollback retires value v and defines a successor v′ (so a successor/compatibility rule exists, Q holds), but the forward path still accepts v for a new act (new use of the retired value is not fail-closed).
Expected code (one): ROLLBACK_FORWARD_FAIL_CLOSED_VIOLATED (RBP-8).
Invariant tested (one): I6 forward fail-closed.
Why no other code competes: a successor rule exists, so RBP-7 (SUCCESSOR_RULE_ABSENT) does not fire (disjoint on Q, file 02 §3.2); identity/history/references/semantics/authority are intact (not RBP-2..RBP-6); the plan is auditable and local (not RBP-9/RBP-10). The unique fault — retired value still usable for new effects — is named only by RBP-8.
PASS/seal/digest possible? No. (This is the exact input that reached PASS under PATCH1; under PATCH2 it matches RBP-8 ⇒ ¬RBP8 false ⇒ PASS impossible, file 02 proof P-B2.)
Result: REJECTED.

5. New fixture XBI-27 (closes B7) — C2 successor schema weakens forward authorization

Input (one): a C2 schema rollback introduces a successor protocol_version whose schema drops or optionalizes a required authority reference for new packets — e.g. drops founding_authority_ref, makes approval_mode optional, or omits approval/quorum/principal refs while approval_mode = APPROVAL_USED (file 05 §2).
Expected code (one): ROLLBACK_WEAKENS_AUTHORITY (RBP-6).
Invariant tested (one): I5 authority non-weakening — forward branch, now covering C2 (file 05).
Why no other code competes: the successor schema exists (so RBP-7 does not fire) and old packets remain readable/resolvable under their original version (not RBP-2/RBP-3/RBP-4) with unchanged old meaning (not RBP-5); the retired schema is not the issue (not RBP-8). The fault is that the forward schema requires fewer authority inputs. RBP-6 (rank 6) is the earliest and only code that names authority weakening, and precedes RBP-7/RBP-8.
PASS/seal/digest possible? No.
Result: REJECTED.

5.1 XBI-27 sub-cases (each single-coded, all → RBP-6)

For self-review coverage (file 09 A8/A9/A10), the C2-weakening input has three governed sub-shapes; each is one input → the one code RBP-6:

27-a: successor C2 schema drops founding_authority_ref ⇒ ROLLBACK_WEAKENS_AUTHORITY.
27-b: successor C2 schema makes approval_mode optional ⇒ ROLLBACK_WEAKENS_AUTHORITY.
27-c: successor C2 schema omits approval/quorum/principal refs while approval_mode = APPROVAL_USED ⇒ ROLLBACK_WEAKENS_AUTHORITY. All three share code RBP-6 because they are instances of one invariant violation (forward I5); they are distinct inputs but not competing codes — this is allowed (one input → one code; several inputs may map to the same code).

6. Consolidated XBI ledger after PATCH2

Fixture	Input (one)	Code (one)	RBP	Status
XBI-11	drop referenced vocabulary value	`ROLLBACK_DELETES_REFERENCED_IDENTITY`	2	retained
XBI-12	delete referenced owner/scope row	`ROLLBACK_DELETES_REFERENCED_IDENTITY`	2	retained
XBI-13A	delete referenced hash identity	`ROLLBACK_DELETES_REFERENCED_IDENTITY`	2	PATCH2 (split)
XBI-13B	hash identity remains, payload erased	`ROLLBACK_ERASES_HISTORY`	4	PATCH2 (split)
XBI-14A	policy ref disabled, C2 ref cannot resolve	`ROLLBACK_ORPHANS_DEPENDENCY`	3	PATCH2 (split)
XBI-14B	C2 ref resolves but policy re-meant	`ROLLBACK_CHANGES_HISTORICAL_SEMANTICS`	5	PATCH2 (split)
XBI-15	approval evidence erased	`ROLLBACK_ERASES_HISTORY`	4	retained
XBI-16	old approval-required envelope weakened	`ROLLBACK_WEAKENS_AUTHORITY`	6	retained
XBI-17	no successor rule	`ROLLBACK_SUCCESSOR_RULE_ABSENT`	7	retained
XBI-18	no audit trail	`ROLLBACK_AUDIT_TRAIL_ABSENT`	9	retained (renum 8→9)
XBI-19A	silent C2 rewrite, meaning preserved	`ROLLBACK_NOT_LOCAL`	10	PATCH2 (split)
XBI-19B	silent C2 rewrite, meaning changed	`ROLLBACK_CHANGES_HISTORICAL_SEMANTICS`	5	PATCH2 (split)
XBI-20	C7 omitted while `APPROVAL_USED`	`C7_REQUIRED_WHEN_APPROVAL_USED`	—	retained
XBI-21	C7 demanded while `APPROVAL_NOT_USED_BY_POLICY`	`C7_OPTIONALITY_CONFLICT`	—	retained
XBI-22	baseline acceptance → open P2	`BASELINE_ACCEPTANCE_NOT_P2_OPEN_AUTHORIZATION`	—	retained
XBI-23	plan reviewed, no Chairman token	`G2_EXECUTION_REQUIRES_SEPARATE_AUTHORIZATION`	—	retained
XBI-24	generic/wrong-scope Chairman token	`CHAIRMAN_AUTHORIZATION_SCOPE_MISMATCH`	—	retained
XBI-25	P2-open plan omits rollback proof	`P2_OPEN_GATE_ROLLBACK_PROOF_MISSING`	—	retained
XBI-26	successor exists but retired value still admissible	`ROLLBACK_FORWARD_FAIL_CLOSED_VIOLATED`	8	PATCH2 (new, B6)
XBI-27	C2 successor schema weakens forward authority	`ROLLBACK_WEAKENS_AUTHORITY`	6	PATCH2 (new, B7)

Every row is single-input/single-code. The RBP-coded rows belong to the closed RBP set (file 02 §1); the dash-coded rows belong to the gate/C7 classifiers (files 05/06 of PATCH1), which compose with — not replace — the RBP set.

7. Oracle totality (classifier-scoped, unchanged posture)

The RBP oracle maps each declared rollback-plan input to exactly one code from the closed set { RUNTIME_MUTATION_REJECTED, ROLLBACK_PLAN_ABSENT, ROLLBACK_DELETES_REFERENCED_IDENTITY, ROLLBACK_ORPHANS_DEPENDENCY, ROLLBACK_ERASES_HISTORY, ROLLBACK_CHANGES_HISTORICAL_SEMANTICS, ROLLBACK_WEAKENS_AUTHORITY, ROLLBACK_SUCCESSOR_RULE_ABSENT, ROLLBACK_FORWARD_FAIL_CLOSED_VIOLATED, ROLLBACK_AUDIT_TRAIL_ABSENT, ROLLBACK_NOT_LOCAL, ROLLBACK_CONTRACT_VALID_FOR_REVIEW } (12 codes). Per caveat C2 this totality is classifier-scoped, not runtime-totality: arbitrary malformed serialization / parser failure / implementation behavior requires its own guards in a future authorized lane.

8. Self-result

All fixtures (XBI-11..XBI-27, including the splits) fail closed; none yields PASS/seal/digest/certificate/authority-token/registration-ready output. No destructive, orphaning, history-erasing, semantics-changing, authority-weakening, unsafe-successor, or non-local rollback can reach ROLLBACK_CONTRACT_VALID_FOR_REVIEW. These codes are classifier labels, not new runtime blockers (no carried blocker is resolved or added). Therefore this file does not stop at RS5B_CLOSEOUT_PATCH2_HOLD_XBI_NOT_SINGLE_CODE. REGISTRATION_HOLD retained.