fix: adjudicate received ParticipantStatus per dimension

[sei-ahouseholder]

• Closes Rejected status updates are dropped silently and all-or-nothing (violates liberal-accept) #2235

Summary

A received ParticipantStatus is now adjudicated per dimension instead of as a unit, so refusing one dimension no longer discards the others and no longer aborts the Seam 1 → Seam 2 emit that drives embargo teardown.

Changes

• vultron/core/behaviors/status/nodes/dimension_filter.py (new): FilterParticipantStatusDimensionsNode adjudicates rm, vfd and pxa independently and publishes a filtered ParticipantStatus in which each refused dimension carries the participant's current value forward. It is read-only with respect to the DataLayer, so it runs as a precondition guard ahead of the guarded commit (CLP-10-006). An update whose accepted portion is indistinguishable from current state is refused in full — nothing appended, no ledger entry committed.
• vultron/core/behaviors/status/add_participant_status_tree.py: wires the filter in as Seam 1's second guard, replacing CheckParticipantRMNotClosedNode. That node's FAILURE was what aborted the enclosing Sequence before StatusUpdateGuard / EmitAddCaseStatusToSelfNode.
• vultron/core/behaviors/case/nodes/lifecycle.py: adds the generic, ID-matched ledger_payload_object_override blackboard key, next to its consumer CommitCaseLedgerEntryNode, so the canonical CaseLedgerEntry snapshots the accepted portion rather than the raw assertion. Opt-in, so the other receive trees are unaffected.
• vultron/core/behaviors/status/nodes/append.py: ResolveAndPersistStatusObjectNode persists the filtered status; ValidateRMTransitionNode accepts an rm value that was refused and carried forward upstream, while keeping its all-or-nothing semantics when the append subtree is used standalone.
• vultron/core/behaviors/sync/nodes/participant_status_effect.py (new): ApplyParticipantStatusFromLedgerNode moves here and gains an RM ratchet — an Announce(CaseLedgerEntry) that would regress a replica's rm on the progress scale has that dimension carried forward at the local value (monotonic visibility). Lateral same-rank moves (VALID ↔ INVALID, DEFERRED ↔ ACCEPTED) are the Case Actor re-adjudicating and are applied unchanged.
• vultron/core/states/cs.py: adds is_monotonic_vfd_forward / is_monotonic_pxa_forward, treating each VFD/PXA component as an independent one-way latch.
• vultron/core/behaviors/status/nodes/rm_validation.py (new): ValidateRMTransitionNode and the now-deprecated CheckParticipantRMNotClosedNode move out of append.py. Both append.py (499) and sync/nodes/effects.py (495) sat within 5 lines of the BTND-07-004 500-line cap, so this fix could not land without splitting them.
• Docs: ADR-0060, spec group RSH-05-001 through RSH-05-008, plus notes/received-status-authorization.md and notes/sync-ledger-replication.md.

Scope

em is deliberately not adjudicated at Seam 1 — embargo state belongs to Seam 2, tracked in #2256. The sender still receives 202 Accepted regardless of outcome (#2255) and there is no outbound refusal message (#2259); per the agreed scope, the refusal is made visible through the canonical ledger, which every participant replicates.

CheckParticipantRMNotClosedNode is retained rather than removed (no breaking API changes), marked deprecated with an explicit "do not wire this back in" note.

Verification

• 6 new regression tests in test/core/behaviors/status/test_partial_accept_participant_status.py: refused rm with accepted vfd/pxa, survival of the Seam 2 emit, the ledger snapshot carrying the accepted rm, an RM.CLOSED participant still advancing vfd, whole-update refusal committing no entry, and the replica-side RM ratchet.
• Full unit suite passes (0 failures); integration suite passes (0 failures).
• black, flake8 (incl. CC gate), mypy (1203 files) and pyright (0 errors, 0 warnings) all clean.
• markdownlint-cli2 clean on all new and modified markdown.

Note for reviewers: the guard→append handoff uses the py_trees blackboard, which is process-global and not cleared between BT executions. Both new keys are therefore written on every tick (with None when inapplicable) and matched by object ID on read. This is recorded as a concern in plan/incoming/learnings/.

🤖 Generated with Claude Code

[sei-ahouseholder]

PR Triage: #2280 — fix: adjudicate received ParticipantStatus per dimension

Linked issues: #2235 (Rejected status updates are dropped silently and all-or-nothing (violates liberal-accept))
Changed files: 26 files — core/behaviors, core/states, specs, notes, docs/adr
CI status: ❌ failing
Merge state: ✅ mergeable (UNSTABLE — failing checks only, no conflicts)
Base branch: main
Needs integration tests: yes

---

Findings

#	Phase	Severity	Description	Outcome
phase5-ledger-snapshot-core-shaped-0	spec-conformance	❌ FAIL	dimension_filter.py:273 dumps the core ParticipantStatus into the canonical ledger payload override instead of as_ParticipantStatus.from_core(filtered) — root cause of both fcvcv Invariant Harness failures (CM-18-006, CLP-07-001, RSH-05-004)	fix-now
phase8-ratchet-result-discarded-0	code-review	❌ FAIL	participant_status_effect.py:220 — the RM ratchet result is logged then discarded when the status already exists in the DataLayer; the un-ratcheted status is appended and the replica's RM regresses (RSH-05-007, SYNC-02-002)	fix-now
phase8-absent-case-status-erases-pxa-em-0	code-review	❌ FAIL	dimension_filter.py:175 — an inbound status with no case_status erases the participant's known pxa/em and commits a ledger entry that RSH-05-005 requires be refused in full	fix-now
phase10-no-snapshot-schema-assertion-0	test-coverage	❌ FAIL	The new suite's shape-agnostic accessors (lines 119-130) let 6/6 pass while CI fails; no test asserts the ledger snapshot's wire schema, the absent-case_status branch, or the pre-seeded-DataLayer ratchet path	fix-now
phase8-closed-to-closed-mislabeled-refused-0	code-review	⚠️ IMPROVE	_rm_is_acceptable (:138) reports CLOSED → CLOSED as a refused dimension, so the WARNING and the audit trail name a refusal where nothing was discarded (RSH-05-002)	fix-now
phase8-noop-path-leaves-stale-bb-keys-0	code-review	⚠️ IMPROVE	dimension_filter.py:294 — the _require_datalayer() early return skips _publish(), leaving both output blackboard keys stale, contradicting the node's own docstring (BT-17-003/004)	fix-now
phase10-no-blackboard-leak-test-0	test-coverage	⚠️ IMPROVE	No cross-run leak test for BB_DIMENSION_FILTER / BB_LEDGER_PAYLOAD_OBJECT_OVERRIDE; the ID match at lifecycle.py:177 is untested in both directions	fix-now
phase10-monotonic-helpers-untested-0	test-coverage	⚠️ IMPROVE	is_monotonic_vfd_forward / is_monotonic_pxa_forward have zero direct unit tests	fix-now
phase6-adr-number-collision-0	adr-check	⚠️ IMPROVE	ADR 0060 is claimed by both this PR and older open PR #2275; neither is on main. Renumber this one to 0061 (+8 referencing files)	fix-now

Total: 4 FAIL · 5 IMPROVE · 0 NEW-ISSUE

---

The lead FAIL, in one paragraph

FilterParticipantStatusDimensionsNode._publish() writes the partial-accept
override with filtered.model_dump(...) on the core model. Reproduced
locally, the two paths diverge:

partial accept (override):  [caseEngagement, caseStatus, context, cvdRole,
                             embargoAdherence, id, published, rm, type, updated, vfd]
full accept  (no override): [@context, caseEngagement, caseStatus, context, cvdRole,
                             embargoAdherence, id, name, published, rmState, type,
                             updated, vfdState]

Nested rm/vfd dimension objects, no @context, no name, no
emConsentState. Every other canonical payloadSnapshot in the project is
wire-shaped, so this one entry is unreadable to the invariant harness
(_missing_fields_in_status_snap, cs_observations_from_snap) — and, more
importantly, to every participant it is replicated to via
Announce(CaseLedgerEntry). It also violates the AGENTS.md rule
"Core→Wire Conversion: Use wire_cls.from_core() — never model_dump() +
model_validate()."

CI baseline

fcvcv Invariant Harness is green on main (run 31632495576) and red here
with 2 failed, 46 passed in 1.85s — the assertions ran, so this is a genuine
regression. Already red on main and not attributable to this PR:
fvcv-handoff Invariant Harness, plus the Demo Integration jobs for
fvv / fccv-extension / fvcv-handoff / fvcv-extension / fcv / fcvcv.
This PR takes fcvcv Demo Integration from 7 → 4 failures.

Cleared

Cross-module refactor fidelity (all moved names re-exported, __all__ complete,
no symbol lost, no test monkeypatches an old path); BTND-07-004 500-line cap
(dimension_filter 393, append 341, rm_validation 280, participant_status_effect
256); CLP-10-006 guard/commit/effect ordering; mkdocs build --strict; notes
frontmatter (NF-06-001/002); commit trailers; PR body format; RM progress-scale
ordering and the ADR-0060 lateral carve-out; cs.py monotonicity logic;
embargo_adherence settability.

---

Triage artifact: .claude/pr-2280-triage.json
Next step: run /pr-execute or /pr-ship to apply fixes.

[sei-ahouseholder]

PR Execute: #2280 — fix: adjudicate received ParticipantStatus per dimension

Fixes applied: 9 findings in 1 commit (8cbfc79e), plus 1 catalog commit (b574e701)
Issues filed: 0
Deferred (awaiting your input): 0
Tests run: unit + integration
CI status after push: ⏳ pending — the new run started for b574e701; the previous full run is analysed below
Base sync: ✅ merged main @ d06619ca — 0 conflicts

---

Fixed

Finding	Commit
phase5-ledger-snapshot-core-shaped-0: adjudicated ledger snapshot was core-shaped	8cbfc79e
phase8-ratchet-result-discarded-0: RM ratchet discarded when the status object already existed locally	8cbfc79e
phase8-absent-case-status-erases-pxa-em-0: an omitted caseStatus blanked pxa and em	8cbfc79e
phase10-no-snapshot-schema-assertion-0: no test pinned the snapshot's wire schema	8cbfc79e
phase8-closed-to-closed-mislabeled-refused-0: CLOSED→CLOSED logged as a rewrite (addressed in part — see below)	8cbfc79e
phase8-noop-path-leaves-stale-bb-keys-0: no-op path left a stale blackboard override	8cbfc79e
phase10-no-blackboard-leak-test-0: no cross-execution leak test	8cbfc79e
phase10-monotonic-helpers-untested-0: _is_monotonic_{vfd,pxa}_forward untested	8cbfc79e
phase6-adr-number-collision-0: ADR-0060 collided with open PR #2275 → renumbered 0061	8cbfc79e

Each of the four FAIL fixes was proven to be a real fix by reverting the source and watching the new test go red: KeyError: 'rmState', 'RECEIVED' == 'VALID' alongside a self-contradicting WARNING log, SUCCESS-instead-of-FAILURE, and a surviving stale override.

One finding addressed only in part

phase8-closed-to-closed-mislabeled-refused-0 proposed making _rm_is_acceptable return True for CLOSED→CLOSED. Not applied — that empties refused, which makes ValidateRMTransitionNode re-reject the status and reintroduces the exact silent drop this PR fixes (#2235), and it lets a pure no-op status be appended and ledger-committed. The audit-accuracy half was applied: the operator WARNING now says blocked dimension(s) … with no change to the asserted value when the recorded value matches the assertion, and reserves rewrote dimension(s) … for a real discard.

---

Tests

Full unit suite (rm -rf devlogs && uv run pytest -q -p no:randomly): 6726 passed, 370 skipped, 2 xfailed, 0 failed, exit 0. Counts are derived from the progress characters because this repo's reporter emits no final count line; there are zero FAILED/ERROR lines and no +++ Timeout +++ marker. The rm -rf devlogs prefix is required per #2274.

Directly on the changed behaviour — 148 tests, all passing:

File	Tests
test/core/behaviors/status/test_partial_accept_participant_status.py	18
test/core/states/test_cs_monotonic_predicates.py (new)	117
test/core/behaviors/status/nodes/test_append.py	10
test/core/behaviors/sync/nodes/test_participant_status_effect.py	3

xfail ratchet clean: the only two XFAILs reference #1991 and #1992, both open.

---

Pre-existing CI failures — tracked in #2233

Run 31656171944 @ d06619ca had six failing jobs. All six are one root cause, and all six fail identically on the exact merged base 06bf60c2 (run 31655077093):

Job	Failure
fcvcv / fvcv-handoff Demo Integration	SvcEngageCaseUseCase failed: TransitionParticipantRMtoAccepted → 422 on trigger/engage-case, then Timed out waiting … {RM.ACCEPTED}; current=RM.VALID
fcvcv / fv / fcv-reject Invariant Harness	test_invariant_5_expected_event_types_present[engage_case]
fvcv-handoff Invariant Harness	same, plus test_fvcv_handoff_vendor2_rm_triage_observed

The harness failures are downstream of the 422: it aborts the trigger before GuardedCommitCaseLedgerEntryBT, so no engage_case entry exists for the assertion #2266 made universal. Causality check against this diff: the 422 is logged ~95s before the first FilterParticipantStatusDimensionsNode line in the same job, so the new guard cannot be upstream of it. The base is in fact worse — six Demo Integration jobs fail there versus two here. Evidence handoff: #2233 comment.

The PR-attributable CI failure is fixed. fcvcv Invariant Harness previously failed the emConsentState/cvdRole snapshot-schema assertions that triage attributed to this PR; it is now 1 failed, 48 passed with only [engage_case] left — direct confirmation of the patch-merge fix.

notes/flaky-tests.md had routed four of these jobs to #2216 and #2121, both closed. Commit b574e701 evicts the stale pointers, adds the two missing rows, and records that this row set is deterministic rather than flaky.

---

Not done

• Review threads: none to resolve — gh api repos/CERTCC/Vultron/pulls/2280/comments returns 0 inline comments; the only review is triage's own.
• NEW-ISSUE findings: none — all 9 triage findings were fix-now.
• Separately noted, not this PR's problem: open PR #2210 claims ADR-0059, which is already taken on main.

---

Execute artifact: .claude/pr-2280-execute.json
Next step: /pr-verify — running now as part of /pr-ship.

[sei-ahouseholder]

PR Verify: #2280 — fix: adjudicate received ParticipantStatus per dimension

Overall verdict: ❌ GAPS-FOUND — blocking flag is ⚠️ UNVERIFIED-CI-FAILING
CI status: ❌ failing — 15 pass, 6 fail @ b574e701
Merge state: ✅ MERGEABLE (UNSTABLE — CI red, no conflict) — base main
Base sync in execute: ✅ merged @ d06619ca (0 conflicts resolved)
Integrity check: ✅ all 9 findings accounted for

---

Finding Verdicts

Every fix was checked at HEAD (b574e701), not just in its commit diff.

Finding	Severity	Outcome	Verdict
phase5-ledger-snapshot-core-shaped-0	❌ FAIL	fixed @ 8cbfc79e	✅ CONFIRMED — _accepted_wire_patch + "fields" present at dimension_filter.py:77,334
phase8-ratchet-result-discarded-0	❌ FAIL	fixed @ 8cbfc79e	✅ CONFIRMED — unconditional self.datalayer.save(status_obj) at participant_status_effect.py:226
phase8-absent-case-status-erases-pxa-em-0	❌ FAIL	fixed @ 8cbfc79e	✅ CONFIRMED — carry-forward branch at dimension_filter.py:232
phase10-no-snapshot-schema-assertion-0	❌ FAIL	fixed @ 8cbfc79e	✅ CONFIRMED — emConsentState/cvdRole/no-snake-twin assertions at test_partial_accept_participant_status.py:427
phase8-closed-to-closed-mislabeled-refused-0	⚠️ IMPROVE	fixed (in part) @ 8cbfc79e	✅ CONFIRMED — rewrote/blocked wording split at dimension_filter.py:429; the _rm_is_acceptable half was deliberately declined
phase8-noop-path-leaves-stale-bb-keys-0	⚠️ IMPROVE	fixed @ 8cbfc79e	✅ CONFIRMED — unconditional self._publish((), None) at dimension_filter.py:353
phase10-no-blackboard-leak-test-0	⚠️ IMPROVE	fixed @ 8cbfc79e	✅ CONFIRMED — TestLedgerOverrideDoesNotLeakBetweenExecutions (3 tests) at :723
phase10-monotonic-helpers-untested-0	⚠️ IMPROVE	fixed @ 8cbfc79e	✅ CONFIRMED — test/core/states/test_cs_monotonic_predicates.py present, 117 tests
phase6-adr-number-collision-0	⚠️ IMPROVE	fixed @ 8cbfc79e	✅ CONFIRMED — docs/adr/0061-per-dimension-partial-accept.md at HEAD, no 0060 blob

No conflict markers anywhere in the tree.

---

⚠️ Why the verdict is not READY-TO-MERGE

pr-verify Phase 3 is unconditional: integration_tests_run == true and the integration CI jobs are not green, so every finding carries UNVERIFIED-CI-FAILING and rule 2 fires. That is the verdict, and it is not being upgraded here. What the evidence actually shows, though, is worth stating precisely:

Job	Failure	Base 06bf60c2?
fcvcv Demo Integration	SvcEngageCaseUseCase failed: TransitionParticipantRMtoAccepted → 422	❌ also fails
fvcv-handoff Demo Integration	same 422	❌ also fails
fcvcv Invariant Harness	test_invariant_5_expected_event_types_present[engage_case] — 1 failed, 48 passed	❌ also fails
fv Invariant Harness	same	❌ also fails
fcv-reject Invariant Harness	same	❌ also fails
fvcv-handoff Invariant Harness	same + test_fvcv_handoff_vendor2_rm_triage_observed	❌ also fails

All six are one root cause — #2233, the engage-case 422 — and all six fail identically on the exact merged base (base run 31655077093 vs this run). The base is worse: six Demo Integration jobs fail there versus two here. Evidence handoff: #2233 comment.

The CI failure that triage did attribute to this PR is fixed: fcvcv Invariant Harness was failing the emConsentState/cvdRole payload-snapshot assertions and is now 1 failed, 48 passed with only [engage_case] left.

Unit suite locally: 6726 passed, 370 skipped, 2 xfailed, 0 failed (exit 0). 148 tests cover the changed behaviour directly. xfail ratchet clean (#1991, #1992 both open).

So the gap is not a defect in this branch — it is that main cannot currently produce a green Demo Integration run, which is exactly what #2251 is waiting on before required status checks can be enabled. Clearing this verdict needs a human call: merge on the documented base-parity evidence above, or land #2233 first.

---

Artifacts preserved — .claude/pr-2280-triage.json and .claude/pr-2280-execute.json remain for a re-run.

[sei-ahouseholder]

PR Execute: #2280 — fix: adjudicate received ParticipantStatus per dimension

Fixes applied: 9 findings (1 commit) + 1 merge resolution commit + 1 test commit
Issues filed: 0
Deferred (awaiting your input): 0
Tests run: unit + integration
CI status after push: ❌ failing — 6 jobs, all pre-existing (#2233)
Base sync: ✅ merged main @ 92935731 — 4 conflicts resolved (+ 1 semantic conflict)

Re-run of execute after the base moved: #2278 (#2232, ADR-0062) landed on main and put this branch into conflict. The 9 triage findings were already fixed at 8cbfc79e and are unchanged; this pass resolved the merge.

---

Conflicts Resolved

Path	Resolution
docs/adr/index.md	List-append false conflict — kept both entries, 0061 before 0062
mkdocs.yml	Same treatment in the nav
vultron/core/behaviors/status/nodes/append.py	Move-vs-edit. main edited ValidateRMTransitionNode / CheckParticipantRMNotClosedNode in place to use read_rm_states(); this branch had already relocated both classes to status/nodes/rm_validation.py for the 500-line cap (BTND-07-004). Kept the move and hand-transplanted all three of main's hunks to the new location, so nothing was lost. The PR's own carry-forward accept (_rm_was_carried_forward, RSH-05) still runs ahead of the terminal RM.CLOSED check
vultron/core/behaviors/sync/nodes/effects.py	Import-set conflict — kept participant_status_rm_state (used by ApplyCloseCaseFromLedgerNode), dropped main's ParticipantStatus name, unused here since this branch moved ApplyParticipantStatusFromLedgerNode into its own module

Semantic Conflict Resolved

Both sides applied cleanly in participant_status_effect.py, and the result was broken.

This branch's _local_rm_state() read the replica's RM floor with getattr(current, "rm", None) and returned None when unreadable — exactly the degraded-default pattern main's ARCH-15-001/ARCH-15-002 sweep removed everywhere it existed. A None floor makes _ratchet_rm() a no-op, so a regressing ledger entry would be applied unchecked (the #2264 failure mode), silently, because the ratchet only logs when it refuses something. main's ratchet test guards _NORMALIZE_WIRE_TO_CORE only, so nothing caught it.

Replaced with read_rm_states(), which distinguishes a shape mismatch (FAILURE) from a legitimate absence — a participant with no recorded status yet has no floor to enforce. Covered by a new test (a28328cf), verified red against the pre-fix accessor and green after.

Left as-is after review: append.py:266 (hasattr(status_obj, "rm") or hasattr(status_obj, "vfd")) is a "is this a ParticipantStatus at all" type check that already returns FAILURE, not a degraded default, so it is ARCH-15-conformant in outcome.

---

Fixed

Finding	Commit
phase5-ledger-snapshot-core-shaped-0: ledger payloadSnapshot must keep the sender's wire shape	8cbfc79e
phase8-ratchet-result-discarded-0: ratcheted status not saved, so the read-back regressed the replica	8cbfc79e
phase8-absent-case-status-erases-pxa-em-0: absent caseStatus erased current pxa/em	8cbfc79e
phase10-no-snapshot-schema-assertion-0: snapshot test asserted values, not key spelling	8cbfc79e
phase8-closed-to-closed-mislabeled-refused-0: CLOSED→CLOSED narrative wording	8cbfc79e
phase8-noop-path-leaves-stale-bb-keys-0: no-op path left stale blackboard keys (BT-17-003/004)	8cbfc79e
phase10-no-blackboard-leak-test-0: no test for override leakage between executions	8cbfc79e
phase10-monotonic-helpers-untested-0: monotonic CS predicates untested	8cbfc79e
phase6-adr-number-collision-0: ADR 0060 collided with open PR #2275	8cbfc79e

On phase8-closed-to-closed-mislabeled-refused-0 only the audit-accuracy half was applied. The reviewer's _rm_is_acceptable() suggestion was declined: making CLOSED→CLOSED acceptable empties refused, so ValidateRMTransitionNode re-rejects the snapshot and #2235 returns — and it would also let a pure no-op status be appended and ledger-committed.

---

Tests

Suite	Result
Unit suite on the merged tree	6805 passed, 370 skipped, 2 xfailed, 0 failed / 0 errors, no timeouts (exit 0)
-m integration subset	exit 0; 3 xfail/xpass, all referencing open issues
xfail ratchet	#1898, #1991, #1992, #1993, #1994, #2150 — all OPEN

Counts are derived from progress characters: this repo's reporter (addopts = "-ra -q") emits no final count line. integration_tests/ collects nothing (exit 5) — it holds a README.md and a demo/ tree, no pytest modules; CI's Demo Integration jobs cover that path.

CI — 6 failing jobs, all pre-existing

fcvcv / fvcv-handoff Demo Integration and all four Invariant Harness jobs fail on one cause, tracked in #2233 (OPEN, Epic #2230):

SvcEngageCaseUseCase failed: TransitionParticipantRMtoAccepted
  → 422 on POST .../trigger/engage-case
  → no engage_case ledger entry
  → test_invariant_5_expected_event_types_present[engage_case] fails
  → Demo Integration times out on "CaseActor reflects <actor> at RM.ACCEPTED"

Byte-identical signature on base run 31655077093 @ 06bf60c2, where six Demo Integration jobs fail versus two here. No read_rm_states / shape-mismatch errors appear anywhere in the failing logs, and Tests (pytest), Lint (mypy) and Lint (pyright) all pass on the merged tree.

The one PR-attributable CI failure is fixed: fcvcv Invariant Harness previously failed on the emConsentState / cvdRole snapshot schema and now fails only on [engage_case] (1 failed, 48 passed).

---

Execute artifact: .claude/pr-2280-execute.json
Next step: /pr-verify.

[sei-ahouseholder]

PR Verify: #2280 — fix: adjudicate received ParticipantStatus per dimension

Overall verdict: ❌ GAPS-FOUND — UNVERIFIED-CI-FAILING
CI status: ❌ failing — 6 jobs
Merge state: ✅ MERGEABLE (UNSTABLE — the failing checks below, not a conflict) — base main
Base sync in execute: ✅ merged @ 92935731 (4 conflicts resolved, + 1 semantic conflict)
Integrity check: ✅ all 9 findings accounted for

The merge conflict that sent this back through the pipeline is resolved. git grep finds no surviving conflict markers, and the live merge-state check is MERGEABLE. The verdict is held open by CI alone.

---

Finding Verdicts

All nine re-verified by reading each file at HEAD (git show HEAD:), not from the commit diff — the merge rewrote two of these files after the fixes landed.

Finding	Severity	Outcome	Verdict
phase5-ledger-snapshot-core-shaped-0	❌ FAIL	fixed @ 8cbfc79e	✅ CONFIRMED — _accepted_wire_patch at dimension_filter.py:77, used at :334
phase8-ratchet-result-discarded-0	❌ FAIL	fixed @ 8cbfc79e	✅ CONFIRMED — unconditional save(status_obj) at participant_status_effect.py:248 survived the merge
phase8-absent-case-status-erases-pxa-em-0	❌ FAIL	fixed @ 8cbfc79e	✅ CONFIRMED — dimension_filter.py:231-235 carries the receiver's case_status forward
phase10-no-snapshot-schema-assertion-0	❌ FAIL	fixed @ 8cbfc79e	✅ CONFIRMED — key-spelling assertions at test_partial_accept_participant_status.py:455-461
phase8-closed-to-closed-mislabeled-refused-0	⚠️ IMPROVE	fixed @ 8cbfc79e	✅ CONFIRMED — rewrote/blocked wording at dimension_filter.py:429-431
phase8-noop-path-leaves-stale-bb-keys-0	⚠️ IMPROVE	fixed @ 8cbfc79e	✅ CONFIRMED — every early return publishes ((), None) (:353, :363, :372, :378, :405)
phase10-no-blackboard-leak-test-0	⚠️ IMPROVE	fixed @ 8cbfc79e	✅ CONFIRMED — TestLedgerOverrideDoesNotLeakBetweenExecutions at :778
phase10-monotonic-helpers-untested-0	⚠️ IMPROVE	fixed @ 8cbfc79e	✅ CONFIRMED — 14 tests in test_cs_monotonic_predicates.py
phase6-adr-number-collision-0	⚠️ IMPROVE	fixed @ 8cbfc79e	✅ CONFIRMED — docs/adr/0061-…md on disk; index.md:132 and mkdocs.yml:280 agree; no 0060 on this branch

---

❌ CI — the only thing blocking

Job	Status
Tests (pytest), Lint (mypy), Lint (pyright), Lint (black), Lint (flake8), Spec Lint, docs-build-check, CodeQL, fv + fcv-reject Demo Integration	✅ pass
fcvcv / fvcv-handoff Demo Integration, all four Invariant Harness jobs	❌ fail

All six failures are the same pre-existing cause, tracked in #2233 (OPEN, Epic #2230): SvcEngageCaseUseCase failed: TransitionParticipantRMtoAccepted → 422 on POST .../trigger/engage-case → no engage_case ledger entry → test_invariant_5_expected_event_types_present[engage_case] fails, and the Demo Integration jobs time out on CaseActor reflects <actor> at RM.ACCEPTED.

Evidence that none of it is this PR's: the signature is byte-identical on base run 31655077093 @ 06bf60c2, where six Demo Integration jobs fail versus two here. No shape-mismatch or read_rm_states error appears anywhere in the failing logs.

The one CI failure that was this PR's is fixed: fcvcv Invariant Harness previously failed on the emConsentState / cvdRole snapshot schema and now fails only on [engage_case] (1 failed, 48 passed).

Verify has no carve-out for pre-existing CI failures — a red check is a red check — so the verdict stays GAPS-FOUND. Merging this PR is a judgment call for a human, since the gap is #2233, not #2280.

Post-merge suite

Run on the merged tree, not the pre-merge branch: 6805 passed, 370 skipped, 2 xfailed, 0 failed, 0 errors, no timeouts. -m integration subset green. xfail ratchet clean (#1898, #1991, #1992, #1993, #1994, #2150 all OPEN).

---

Artifacts preserved — re-run /pr-execute to address gaps (here: none actionable in this PR; the gap is #2233).