Zero new failures, and every remaining failure owned by a named task
A change lands with ZERO NEW failures, and every failure that remains is
owned by a named task (user, standing policy; restated by the coordinator for
LAND-OC-FIX R4, 2026-10-07). The workspace’s suite gates (ws validate, the
engine and the web ledgers alike) carry it out as ONE judgement over the
target’s recorded run, keyed by its tree:
- a failure the target’s run does not have is NEW and fails the gate — even in a file a task owns, and even when the test would pass alone (nothing is re-run: UR-17);
- a failure that persists from the target but fails differently (another digest) is a new defect under an old identity and fails the gate;
- a failure that persists from the target the same way passes only when an
OPEN task row of
docs/trackers/TASKS.md(- [ ] **<ID>**, with its continuation lines) names its test’s repository path (engine/test/…,tests/…); the gate names the owner. One no open row names fails the gate.
A BUDGET breach joins the same judgement through the coordinator’s amendment UR-24-BUDGETS (by: coordinator, 2026-10-07). That clause is the coordinator’s, not the user’s.
A test that SKIPS did not pass: the engine runner names every dispatched skip
(<file>::<name>, with its reason), and the skip is judged like a failure
(<file>::<name> (skipped), its digest the reason). An unnamed skip, a
withheld test, a todo or a cancellation stays an independent refusal.