09 - BLOCKER H: Author / Rehearse / Seal Order (in-blueprint)

Codex recheck-2 finding (doc 11)

The package sequence was unusable for complete authoring/rehearsal: PKG-B authored seed DATA and PKG-C rehearsed/sealed before PKG-D authored #11/#20/#26/#27 (legacy closure, gateway, repoint, rollback bindings); the normative S12 seal preceded S13 repoint authoring even though #26/#27 are child-contract rows that must exist before a complete manifest seal. Required fix: move all #11/#20/#26/#27 authoring and denominator/classifier/rollback bindings before the complete rehearsal seal; rehearse exactly the same complete artifact set later applied by PKG-E; separate authoring, rehearsal seal, production seal/stage, and live cutover without reordering numbered normative dependencies; add a guard.

The fix — strict author → validate → rehearse → hash → seal → review → authorize → apply order

1. Author all load-bearing artifacts first (PKG-B). PKG-B now authors not only the catalog/27-item rows / #20 / #21 / gates / hashes / thresholds, but also #11 reverse-closure, the U_legacy denominator, #26 gateway identity pin, and #27 repoint old/new source + rollback bindings — before any rehearsal seal. (The blocker-C legacy-disposition rows are NOT authored here; they are fail-closed pending the amendment.)
2. Static validation of the authored artifacts against the approved byte DDL.
3. Rehearse the EXACT complete set (PKG-C) inside BEGIN..ROLLBACK — including #11/#26/#27 and the denominator — so the rehearsed/sealed artifact set is byte-identical to what PKG-E later applies.
4. Compute seal hashes from the final authored artifacts, then seal the complete set.
5. PKG-D becomes read-only RE-VALIDATION + the Codex re-audit — it authors nothing the rehearsal seal did not cover; it re-confirms the sealed #26/#27 source hashes still match live and that #11 closure (legacy_reached = 0) still holds.
6. Production seal/stage (PKG-E) and live cutover (PKG-F/G) are distinct later phases; activation never precedes seal/review/authorization.

New guard

G-SEAL-AFTER-AUTHOR-REHEARSE (doc 06) FAILS if any manifest is sealed before the exact artifact set it seals has been authored and rehearsed, if seal hashes are not computed from the final authored artifacts, or if any activation precedes seal/review/authorization.

Where patched

• Blueprint doc 07: PKG-B scope expanded to author #11/#26/#27 + denominator; PKG-C rehearses/seals the complete set; PKG-D demoted to read-only re-validation; the package-sequencing notes updated.
• Blueprint doc 04: the "Author/rehearse/seal order" dependency note added (S13 authoring before the complete rehearsal seal).
• Blueprint doc 06: G-SEAL-AFTER-AUTHOR-REHEARSE added.

Dependency on blocker C

The ordering is fully fixed in-blueprint. The completeness of the sealed set still depends on the blocker-C amendment for the legacy-disposition rows: until the design owner rules (Option α adds a typed contract; Option β collapses to approved primitives), those rows are not authored, so the complete-authoring readiness gate (AUTHORING_PLANNING) cannot be declared green. The order is correct; the set is complete only after C resolves.

PG-native, no new surface

Pure reordering of the package/step sequence plus one test/verification guard. No new surface/column/family/hash.

Self-check

PASS only if no seal precedes the artifact it seals, the rehearsal covers the exact applied set, and the phases are separated. PASS for the ordering (PKG-B authors #11/#26/#27 + denominator before the PKG-C rehearsal seal; PKG-D read-only; G-SEAL-AFTER-AUTHOR-REHEARSE). Full authoring readiness remains gated on the blocker-C amendment.