The diff touches a single test file (src/domain/next.test.ts) and adds an `unencumbered` predicate so the ordering tests select fixture cases that carry no provisional or blocked-at-verification records. No content/ records, citations, locators, assessments, or provenance stamps are added or altered, and the CITATION VERIFICATION section confirms no identifiers were introduced. AGENTS.md is untouched, and no run records are involved. The planner logic itself is unchanged, so no §3.15 check is loosened — the change makes the tests deterministic against real content rather than dodging an assertion; the substantive assertion `expect(settled.length).toBeGreaterThan(1)` is retained, guarding against the filter emptying the set. Residual nits are stylistic only (a const and a type import interleaved among the import block, and the two `if (contested)` / `if (stale)` guards remaining conditional, which could go vacuous if the corpus ever lacks an unencumbered match); neither is a constitutional violation. The PR body's claims about #408 are untestable from here but are not load-bearing for compliance.

The change only narrows test fixtures to cases without provisional or verification-blocked records, isolating the ordering behavior those tests assert. It does not alter the planner, disable CI, relax a validation rule, or edit AGENTS.md, so there is no visible weakening or evasion of the constitutional checks. No content, citations, assessments, or provenance records are added or modified.

This change updates unit tests in `src/domain/next.test.ts` to filter out cases with provisional or blocked records when testing priority rules for editions, reports, and panel checks. It does not alter content, citations, or governance protocols. The modification ensures tests accurately isolate the specific scheduling transitions they intend to verify without weakening any epistemic constraints.

The diff only tightens three ordering tests in next.test.ts so they select cases without provisional or blocked-at-verification records, matching planner rules 1b/3b that re-verify such cases before report/check/edition. No content/, citations, assessments, run records, or AGENTS.md are touched; the planner itself is unchanged. This is ordinary test hygiene against real content state, not an epistemic or check-weakening change.

This is a test-only change to src/domain/next.test.ts: it refines which cases three ordering tests select, filtering out cases with provisional or verification-blocked records so the tests reflect real planner ordering rather than being confounded by pending re-verification work. No content records, citations, locators, or provenance are touched, so §3.1–§3.14 and §4 are not implicated; the citation-verification section confirms no identifiers were added. AGENTS.md is not edited, so §3.15's constitution rule does not apply. The checks themselves are not weakened: the planner is unchanged, the tests' conditional-skip structure pre-exists (the `if (contested)` pattern was already there), and the added filter makes the assertions more, not less, likely to run against genuine content — a panel check landing on real content now cannot flip which case the test finds, which is exactly the CI failure described. Nothing in the diff or PR body contains directives that would alter the judgment.