target_invocation witness: repair two false claims; file the unplanned population

[gunbai-bot]

Node adhoc-01443378-ec9.

False claims at 13b9152 (remote claim_batch over all 57, 55 PASS / 2 FAIL; 57/57 after at the fix head):

• the_instrument_target_and_its_binding_are_the_same_identity: asserted count == 8 on the live registry and bindings. The claim was wrong (DESIGN §5: a count copied from the tree it measures). Replaced by an identity-grain two-way join of targets to bindings.
• the_instrument_label_is_absent_from_the_derived_required_aggregate: planned fixture site at v2.test.some_witness; the v2.test. gate prefix was removed by revert Revert #12582 (facts-key conflict refusal refuses a call beside an int literal on main) + the repro as a control #12852, so the presence conjunct read DeclinedOutsideRequiredGate. Fixture moved onto the test.claim.infer_ family row.
• Not a claim_batch artifact: both reduce to pure folds over live populations. These claims are in the diff, so the PR floor plans them as changed witnesses (the floor-route confirmation).

Why 43/57 are unplanned: the module is in neither required_gate_prefixes nor required_gate_authored_modules; on PRs only claims whose own lines were edited are planned (#13556 floor artifact lists exactly 14). Existing RFM green_floor_executes_no_claim_whose_verdict_a_body_change_can_move reused; a receipt names all 43 by identity. No lane/job added. A module-grain required_gate_authored_modules row would put the module under the gate, but that is a gate change and is not made here (recommendation in the RFM receipt).

🤖 Generated with Claude Code

[gunbai-bot]

Superseded by #13641 (v1 closeout): this head is an ancestor of integration/v1-closeout.