V2-LENS-0: dispatch trigger, checkpoints, and permanence laws

dag/gunbc/roadmap_authority.dag

@@ -1039,8 +1039,8 @@ fn declared_roadmap_nodes() -> List<RoadmapNode> {
         headline: "V2-LENS-0: every registered v2 lens executes over one exact corpus subject through the CI2 route",
         boundary: "Starts after CI2-0 terminal acceptance (operator convergence mandate 2026-08-09). Roster derived from the existing v2 lens registry — no second hand-maintained suite list; every registry identity appears exactly once in the plan; every applicable lens body executes through the v2/CI2 route with ordinary v1 interpreter lens executions zero. Each lens returns a reading or a typed located refusal — Unavailable retains its exact trigger and counts as executed, never as accepted capability. One exact whole-corpus fact pack per process; a repeated query constructs nothing.",
         displaced_cost: "Registry completeness proves metadata coherence, not execution: many LensContract rows carry NoConsumerWitness, and an inert lens is coverage by illusion. Binary acquisition, not lens logic, was the recurring execution failure; #8054 proved a whole-corpus query executes and its expense is corpus-fact construction, not lens evaluation.",
-        first_slice: "Derive the complete roster from the registry; execute over one exact subject via the CI2 plan operation; aggregate one LensSuiteReceipt: registered, uniquely planned, executed, accepted readings, expected and unexpected refusals, unaccounted, fact-pack identity and construction count, v1 executions, fallback, preparation and execution walls.",
-        red_control: "Registered-but-never-invoked, wrong-subject invocation, a missing input becoming an empty successful reading, and an invented exact answer on an unavailable host fact all refuse. Existing lens-family RED controls stay discriminating. Fact-pack construction count is one per process and a second query constructs zero. Warm execution after fact construction within five seconds; cold construction measured separately.",
+        first_slice: "Dispatch trigger (operator 2026-08-09): one ordinary NON-FIXTURE test declaration completes canonical parse-normalize-resolve-infer-realize-execute in #8043, with the removed parser diagnostic named, a planted nearby syntax RED still refusing, and the roster identity joined. Checkpoint A: one deliberately boring lens selected MECHANICALLY from the registry (deterministic, hermetic, no host effect, no base-revision graph, small fact input) — a checkpoint, never a completed capsule. Checkpoint B: derive the complete roster, execute all through one shared fact pack, one LensSuiteReceipt. Checkpoint C: enroll the suite permanently in gunbc test.",
+        red_control: "Permanence laws: registering a lens AUTO-ENROLLS it (a registry row without an executable body reds gunbc test; a deleted lens leaving a row reds the roster join); executing zero lenses can never produce a successful suite verdict; registered = planned = terminal outcomes with zero duplicates and zero unaccounted; ordinary v1 lens-body executions and fallback are zero. A missing input yields Partial, Unavailable, or Refused — never an empty success. Fact-pack construction is once per process, a second query constructs zero, warm execution within five seconds. The delivery/consumption question returns as ONE lens here (advisory until declaration grain), never a standalone workflow.",
         out_of_scope: "No new universal lens ontology, no separate lens scheduler, no report-only successor PRs. Host or v1-produced input facts may remain explicit temporary bridges but may not decide a lens answer.",
         handback: "One LensSuiteReceipt on one exact corpus subject with zero v1 lens-body executions and at least one real downstream consumer — initially gunbc test — reading it."
       )