F-2 Rollback-Only Coverage Dry-Run — 08 Final GO/NO-GO
08 — Final GO / NO-GO for the Next Step
8.1 This mission's verdict
F-2 rollback-only dry-run: PASS. Inside a single BEGIN..ROLLBACK, the coverage path was wired onto the real collection_registry source as live view objects, seeded with a transient draft ruleset, candidate scan-run + state, and ownership rows producing direct + inherited + missing owner behavior, then exercised for gap, conflict, and six integrity rejections — and rolled back with proven entry==exit from two independent channels. No COMMIT. Zero persistent mutation. The activation gate (os_proposal_approvals=0) is intact.
8.2 Completion-criteria scorecard (mission §7)
| # | Criterion | Result |
|---|---|---|
| 1 | State recovery complete | PASS |
| 2 | Preflight passes | PASS (GO, doc 01) |
| 3 | F-2 runs inside BEGIN..ROLLBACK | PASS |
| 4 | No COMMIT | PASS |
| 5 | Meaningful seam/gap/owner logic tested | PASS (S1–S12) |
| 6 | Corrected thresholds used | PASS (18 / 210 per F-DR-1) |
| 7 | ROLLBACK completed | PASS |
| 8 | Entry==exit proven | PASS (both channels, Δ=0) |
| 9 | KB package published/readable | PASS (doc 09 / verification) |
| 10 | No unsafe mutation occurred | PASS |
⇒ PASS (all 10 met; not PARTIAL — no objective skipped for safety; every F-2 test objective 1–10 achieved).
8.3 Staircase position
| Step | State after F-2 |
|---|---|
| F-0 readiness/boundary design | done |
| F-1 read-only dry-run | done |
| F-2 rollback-only dry-run | done — this package |
| F-3 rollback rehearsal of seam wiring + acyclic containment screen | eligible (F-2 clean) — optional consolidation step |
| F-4 ratified seam wiring (COMMIT) | NO-GO — L2+L4 ratification |
| F-5 small seeded run | NO-GO — F-4 + seed authorization |
| F-6 event-type activation | NO-GO — ratification + per-type auth |
| F-7 first real emit | NO-GO — production activation |
8.4 GO — next executable step
The read-only (F-1) and rollback-only (F-2) paths are both proven. The remaining engineering dry-run value is small; the binding constraint is now authorization, not code. Two non-exclusive GO options:
- GO-A (optional): F-3 consolidation rollback rehearsal — a single
BEGIN..ROLLBACKthat also (a) wires the containment seam to a real acyclic source candidate (pivot/IU parent refs) and screens it for cycles, and (b) runs the gap path at a slightly larger subset (e.g. 5–8 collections) to confirm bounded behavior toward the 210 ceiling. Still zero COMMIT. This hardens the F-4 wiring spec before any ratified commit. - GO-B (recommended primary): GPT review of this F-2 package + human B-RATIFY intake for the authorization model. This is the only path that unblocks F-4+.
8.5 NO-GO (unchanged, firm)
- F-4 seam COMMIT and all activation (F-4..F-7): NO-GO — blocked by the unsolved L2 council + L4 sovereign ratification of the authorization model (
os_proposal_approvals=0). Agent cannot satisfy; human-only. - Any axis/assignment/topic coverage dry-run: NO-GO —
axis_registry/axis_assignment/coverage_ruleABSENT. - Object-grain (
birth_registry) dry-run: NO-GO for a first run — ~10⁶ source; needs SB-13 cursor. - Any emit /
system_issues/ notification / DOT / worker start: NO-GO — zero-emit boundary.
8.6 Remaining blockers
- L2+L4 ratification of the authorization model (the one gate) — human B-RATIFY intake. Unsolved.
- Ratified seam-wiring decision (collection-grain inventory; acyclic containment source) — governance decision, with the F2-1
::textcast folded into its DDL. - Axis substrate build — unblocks axis coverage.
- F2-6 — reconcile/locate the missing GPT-review artifact(s).
- F2-7 — restore the canonical governing-law file.
8.7 Next macro
ONE_ROOF_PHASE1_COVERAGE_DRY_RUN_F3_CONSOLIDATION_ROLLBACK_ONLY (optional GO-A) ∥ GPT review of this F-2 package ∥ human B-RATIFY intake for the authorization model (the only unblock for F-4+). No production activation in any branch.
8.8 Final verdict
GO (F-3 optional consolidation rollback-only / GPT review / human B-RATIFY) · NO-GO (F-4+ activation, axis, object-grain, emit). The read-only and rollback-only coverage paths are proven end-to-end against the real source with zero mutation; the activation boundary and ratification gate are preserved exactly.