Fit identity_cast_emission_route claims in the new-witness eval-step budget

[gunbai-bot]

Summary

• Gated identity-cast claims now include supplied-atom coercion_cast_crossing rows (Identity vs NoTargetCandidate at the diagnostic head) plus inhabitance that still runs emit_closure_from_ingest_located on every required run.
• Native host-run evidence stays off-gate; the duplicate native refusal inhabitance is gone.
• No cost-drop member remains for this subject: gunbc#13489 required floor run 37522243686 (job 112483273531, artifact required-floor-claim-cost) billed every gated identity under the 72300 new-witness budget, which is identity_cast_route_new_witness_eval_step_cost's restoration trigger. That drop and floor_eval_step_cost_drop_identity_cast_route_rows are deleted.

Per-identity billed eval_steps (run 37522243686, required_floor_claim_cost.tsv, head b2ef02e6d7)

identity	eval_steps	vs 72300
an_admitted_identity_cast_emits_through_the_closure_route_holds	11	under — no cost-drop row
a_cast_with_no_witness_refuses_before_realization_holds	11	under — no cost-drop row
an_int_identity_cast_is_an_identity_crossing_holds	302	under — no cost-drop row
an_int_to_bool_cast_has_no_witness_holds	388	under — no cost-drop row

emit_host_identity_cast_native was not planned on that run (unchanged-witness / off-gate).

Test plan

• Required floor on b2ef02e6d7: all four gated identities pass; none over 72300 (run 37522243686).
• Required floor on this head after the cost-drop retirement.
• Native module as a changed witness still builds fn f(x: i32) -> i32.

[gunbai-bot]

review 77130 (on 8c664ae) asked for one of two things: keep a real-path claim on required runs with an operator-approved budget, or declare this as a §4b(3) drop. The brief forbids a cost-debt row and forbids raising the new-witness budget; the emit door measured 657428 eval steps / 1835 cpu_ms on gunbc#13480 run 37486290875, which is also over the 100ms new-witness envelope, so it cannot stay enrolled as a required-run claim.

c4ee5b1 takes the second arm the review named:

• Gated tests are renamed to what they judge (an_int_identity_cast_is_an_identity_crossing_holds, an_int_to_bool_cast_has_no_witness_holds) so the old names are not a meaning-fork.
• emit_closure_from_ingest_located inhabitance stays on v2.test.execution.emit_host_identity_cast_native.
• That merge-time loss is gunbc.rung_drop identity_cast_closure_inhabitance_off_the_required_gate, with population in v2.workflow.floor_eval_step_cost_drop floor_eval_step_cost_drop_identity_cast_closure_inhabitance_rows. When those native claims run as changed witnesses they still execute; a semantic red or wall-clock crossing still blocks; only the eval-step overrun is reported.

That is the same shape as python_to_typescript_compile_inhabitance_off_the_required_gate, not a silent hole.

— sent from lively-boar-148

[gunbai-bot]

review 77172 is right: identity_cast_route_new_witness_eval_step_cost is retired only by its trigger, and relocating the door off the required gate is one of the moves that trigger names as not restoration (DESIGN §4b). 85bb90b restores an_admitted_identity_cast_emits_through_the_closure_route_holds and a_cast_with_no_witness_refuses_before_realization_holds on the gate under that drop (measured at the admitted identity on runs 37486290875 / 37493248052). The supplied coercion_cast_crossing rows stay. Native remains host-run evidence, not the last merge-time execution of emit_closure_from_ingest_located.

— sent from lively-boar-148

[gunbai-bot]

review 77178: deleted a_cast_with_no_witness_is_refused_on_the_closure_route_holds from the native module (42029c7). It was the same call and assertion as the gated a_cast_with_no_witness_refuses_before_realization_holds, and it paid refused_cast_route_verdict again without a host build.

— sent from lively-boar-148

[gunbai-bot]

review 77377: regenerated docs/design-rung-drops.md so identity_cast_route_new_witness_eval_step_cost and floor_eval_step_cost_drop_identity_cast_route_rows are gone from the committed roster.

— sent from lively-boar-148

[gunbai-bot]

review 77691: regenerated and committed docs/design-rung-drops.md from gunbc.rung_drop. The standing roster no longer lists identity_cast_route_new_witness_eval_step_cost or floor_eval_step_cost_drop_identity_cast_route_rows.

The rebase skip of the earlier regen commit was the generated-artifact concurrent-divergence driver; this commit is a post-rebase derivation on the merged authorities, not a hand-resolution of the merge.

— sent from lively-boar-148