The diff touches only src/domain/next.test.ts. It gives one scheduler test a 90-second timeout through vitest's options argument and adds a two-line comment. The test's name, its setup, and its visible assertion (toMatchObject on case and verb) are unchanged. Nothing under content/ is added, so there are no citations, locators, or provenance stamps to judge, which matches the verification tool finding no identifiers. The test covers chooseNext's forced-choice path, which is not one of the §3.15 fail-closed checks (schemas, dangling IDs, provenance completeness, real citations). A longer timeout also loosens nothing: a wrong result still fails, and a real hang still fails at 90 s, so this is neither check-weakening nor a reclassification to dodge a check. AGENTS.md is not edited. The comment's date (2026-09-30) is today and agrees with the PR body, which honestly records the better fix (loading cases once) as not done, as §9 asks when it says to report limitations. The change is the smallest coherent fix for a flake that blocked an unrelated PR, and it keeps §8's must-pass test suite reliable without hiding any failure mode it asserts.

The diff changes only the timeout for one existing test from the default limit to 90 seconds; it does not alter, skip, or weaken the test’s assertions. Allowing a known data-loading test more execution time is not a dodge of the fail-closed checks in §3.15, because the same forced-choice and ledger behavior remains tested. No content, citations, provenance records, assessments, or constitutional text are changed. The change is consistent with §8’s requirement that tests pass while retaining coverage of important behavior.

The change solely increases the execution timeout of an existing test in `src/domain/next.test.ts` from default to 90 seconds to prevent flaky test failures on slow CI runners as the case ledger grows. No test assertions are modified or deleted, and the fail-closed CI checks remain fully intact. The modification touches no content files, citations, or constitutional governance files.

The diff only raises one vitest timeout from the default to 90_000 ms and adds a comment explaining CI slowness from repeated full-case loads; nothing under content/, no citations, no AGENTS.md edit, no provenance or assessment stamps, and no check-weakening. The change is ordinary test hygiene that does not touch any §3 epistemic rule.

This is a routine test-infrastructure change: it raises one test's timeout from twenty seconds to ninety and adds an explanatory comment, changing nothing about what the test asserts. The checks that §3.15 forbids weakening — fail-closed CI for schemas, dangling IDs, provenance completeness, and real citations, plus the ratification gates — are untouched; a timeout extension on a domain test is not one of those checks, and the test still runs to completion and fails if its assertions fail. The diff touches no content files, adds no citations, edits no AGENTS.md, and carries no run records, so no fabrication, confidence, or provenance concern arises. The PR body's candid 'Not done here' section honestly notes the better fix (loading cases once) was deferred — an ordinary scope limitation, not a constitutional defect.