11 — Entry / Exit Hash Proof

Claim: 0 live mutation. Every structural fingerprint captured before the rehearsal equals the fingerprint captured after the rehearsal (and after the one real-function CREATE OR REPLACE … ROLLBACK test).

Structural fingerprint comparison

Key	Entry	Exit	Match
fn_birth_registry_auto md5	1f729b3571a74963089bb3ef388217f3	1f729b3571a74963089bb3ef388217f3	✅
fn_birth_gate md5	b6700aa8c2c5a5f4255f3b119d7dd673	b6700aa8c2c5a5f4255f3b119d7dd673	✅
fn_birth_registry_auto_id md5	3f3515dedf1d5bb0f0f66711297b5b5c	3f3515dedf1d5bb0f0f66711297b5b5c	✅
birth_registry constraints md5	bf3698d7d8d9ac24afb1735f38e95ede	bf3698d7d8d9ac24afb1735f38e95ede	✅
dot_tools trigger-set md5	c77f1a4c62d5ee3480b09280475058b3	c77f1a4c62d5ee3480b09280475058b3	✅
birth_admission_permit table	ABSENT	ABSENT	✅
permit/finalize temp functions	0	0	✅
dot_tools unborn	0	0	✅

Live BLOCK dimensions (unchanged)

Dimension	Entry	Exit	Match
orphan_critical_active	59	59	✅
phantom_real (REAL_MISSING)	6	6	✅
fs FILE_NO_REGISTRY	16	16	✅
dot_pivot_update	1	1	✅

In-rehearsal sub-proofs

• Workstream A–B–D–E ran in one BEGIN … ROLLBACK using ON COMMIT DROP temp tables and pg_temp functions. Post-rollback: to_regclass('public.birth_admission_permit') → NULL; permit/finalize function count → 0.
• Workstream C replaced the real fn_birth_registry_auto inside BEGIN … ROLLBACK: OID stayed 39232, 166 triggers stayed bound, patched conflict target verified present in-txn, then ROLLBACK → md5 returned to 1f729b35… (identical to entry).

Conclusion

Live mutation: NONE. Entry state == Exit state. The production database is byte-identical (structurally) to its pre-rehearsal state. The rehearsal proved the model entirely inside reversible transactions.