import deletion

.gitattributes

@@ -144,6 +144,7 @@ src/v1/stage0/src/std_pareto.rs merge=generated-artifact
 src/v1/stage0/src/std_process_termination.rs merge=generated-artifact
 src/v1/stage0/src/std_realization_schedule.rs merge=generated-artifact
 src/v1/stage0/src/std_reference_binding_observation.rs merge=generated-artifact
+src/v1/stage0/src/std_reference_derived_closure.rs merge=generated-artifact
 src/v1/stage0/src/std_roster_frontier.rs merge=generated-artifact
 src/v1/stage0/src/std_serialization.rs merge=generated-artifact
 src/v1/stage0/src/std_source_annotation.rs merge=generated-artifact
@@ -205,11 +206,15 @@ src/v1/stage0/src/v1_std_core.rs merge=generated-artifact
 src/v1/stage0/src/v1_test_non_ascii_perf_fixture.rs merge=generated-artifact
 src/v1/stage0/src/v1_tests_claim_caret_parse_smoke_test.rs merge=generated-artifact
 src/v1/stage0/src/v1_tests_claim_dag_parse_continuation_operator_witness_test.rs merge=generated-artifact
+src/v1/stage0/src/v1_tests_claim_import_deletion_bridge_witness.rs merge=generated-artifact
 src/v1/stage0/src/v1_tests_claim_namespace_reference_derived_closure_production_admissions_witness_test.rs merge=generated-artifact
 src/v1/stage0/src/v1_tests_claim_occurrence_binding_parser_walk_witness_test.rs merge=generated-artifact
 src/v1/stage0/src/v1_tests_claim_occurrence_identity_debt_receipt_test.rs merge=generated-artifact
 src/v1/stage0/src/v1_tests_claim_ordinary_frontend_observation_test.rs merge=generated-artifact
+src/v1/stage0/src/v1_tests_claim_parser_declaration_category_matrix_witness_test.rs merge=generated-artifact
+src/v1/stage0/src/v1_tests_claim_parser_declaration_matrix_support.rs merge=generated-artifact
 src/v1/stage0/src/v1_tests_claim_pattern_binder_declaration_node_test.rs merge=generated-artifact
+src/v1/stage0/src/v1_tests_claim_reference_derived_closure_witness_test.rs merge=generated-artifact
 src/v1/stage0/src/v1_tests_claim_required_expr_newline_continuation_test.rs merge=generated-artifact
 src/v1/stage0/src/v1_tests_claim_v1_annotation_binding_test.rs merge=generated-artifact
 src/v1/stage0/src/v1_tests_claim_v1_annotation_capture_test.rs merge=generated-artifact

dag/gunbc/ci_spec.dag

@@ -258,8 +258,11 @@ data gunbc_ci_floor_batch_clamp_params: List<RunnableBatchClamp> = [
   gunbc_ci_floor_positional_clamp(overhead_seconds: 540, per_unit_ms: 0),
   gunbc_ci_floor_positional_clamp(overhead_seconds: 120, per_unit_ms: 0),
   gunbc_ci_floor_positional_clamp(overhead_seconds: 125, per_unit_ms: 0),
+  gunbc_ci_floor_positional_clamp(overhead_seconds: 240, per_unit_ms: 1000),
 ]
 
+data gunbc_ci_v1_envelope_known_red_probe_clamp_basis_note: String = "V1-ENVELOPE KNOWN-RED PROBE ROW (the last positional row, added 2026-08-14 on PR #8115): enrolling a row in gunbc.explicit_witness_admission known_red_probe_v1_envelope_roster makes v1_claim_scoped_known_red_probe_batch_enrolled true, which appends one RunnableDiscoveryBatch to the schedule. It is a discovery batch, not a RunnableScopedWitnessBatch, so it carries no scoped clamp of its own and falls to the positional table — the ENROLLMENT COUPLING clause in v2.workflow.ci_floor_plan v1_claim_scoped_known_red_probe_batch_note names exactly this obligation, and it is a fail-closed refusal at arm time rather than a mis-clamp. THE ENROLLMENT THAT GREW THE SCHEDULE, named because this family's template asks for it: src/v1/tests/claim/occurrence_binding_parser_walk_witness_test.dag :: parser_sibling_match_arm_pattern_binder_holds, admitted QuarantineProbeExpectRed because a match-arm pattern binder is stamped ParsedOccurrenceDeclaration with the arm's own ancestors, making the binder a SIBLING of the arm body — structurally red from authored source, not a defect this PR introduced. BASIS, and its honesty class: 95.6s measured wall for the enrolled single row (claim_batch --source-root dag --source-root src/v1 on that entry and function, local x86, cold process). That is NOT a fleet observation — the family's constants are denominated on the arm64 srv host class, and this host is not it — so the 240s overhead is deliberately ~2.5x the local measurement rather than the family's usual 1.6x spread factor, funding the unmeasured host-class difference explicitly instead of pretending the local number is a fleet number. It sits well under v1_claim_scoped_witness_batch's 480s, which shares the dag plus src/v1 envelope over a larger roster, so the ordering between the two rows stays sensible. WHAT THE MEASUREMENT ACTUALLY SPENT: 25.4s of the 95.6 is load_bare_edge_index on one pool entry — the whole-tree bare-reference edge index, a corpus-denominated fixed cost this batch pays like every other resolve, owned by the assembly-decomposition lane and not by this row. per_unit_ms is the signed family witness coefficient 1000, not 0, because the roster is a list that grows: a fixed overhead would silently absorb the second and third enrolled probe. DISSOLVE-ON: the roster empties (the note it enrolls under retires when closure-independent binding makes the dag plus src/v1 envelope unnecessary), at which point the batch unenrolls and this row is deleted with it."
+
 data gunbc_ci_compile_clean_clamp_note: String = "COMPILE-CLEAN LEG CLAMP (prelude coverage, first slice): discharges the compile-clean portion of gunbc_ci_floor_batch_clamp_note FOLLOW-UP ROW (a) — the eager compile-clean install ran OUTSIDE every batch budget and could only red at the step cap, which is how a 15m32s green gate grew past 49 minutes across #7398/#7438 with nothing naming the growth (typecheck-perf investigation, PR #7490). Same shape as the batch clamps: clamp_ms = overhead_seconds*1000 + closure_units * per_unit_ms, where closure_units is the compiled closure's MODULE COUNT — a runtime fact of the scope disposition (whole-tree ~2725 today, affected-set proportional to the diff), so the clamp re-denominates with scope exactly as batch clamps re-denominate with the affected set. Over-clamp is FLOOR-COMPILE-CLEAN-OVER-BUDGET: a typed, located walk refusal (admission grain — the compile receipt's ok is untouched, per the signed admission/verdict split in gunbc_ci_floor_batch_wall_budget_note); never a rerun, never a widen. RED-control hook GUNBC_FLOOR_COMPILE_CLEAN_BUDGET_TIGHTEN_MS lowers the computed clamp (min), never raises. GROWTH DETECTION at row grain is NOT this clamp's job (the family margin ruling, 2026-07-25): per-row 2x drift against dated basis rows rides gunbc.witness_row_cost's comparator over dag/gunbc/compile_clean_cost_basis.tsv — the pass + module_typecheck rows of target/floor-compile-clean-cost-receipt.tsv, basis rows citing arm64 fleet run ids only, BasisAbsent counted until seeded from a cited run's own receipt FILE (the #7475 seeding pattern; an empty basis must count loudly, never read as no-drift). SEEDING SOURCE CORRECTED: this clause used to name the [compile-clean-cost] log lines, because the receipt file had no way off the runner and every row was mirrored into the step log to compensate — ~1,690 lines per floor run, and the same shape cost 8,614 more on the disclosure receipt. The ci job now uploads the receipts as the floor-receipts artifact (gunbc.ci_workflow ci_floor_receipts_upload_step), so seeding downloads the TSV instead of scraping a log, and the writers no longer mirror their bodies. The log lines are GONE, not merely deprecated: a seeding attempt that greps a run log for them will find nothing, which is why this clause is corrected rather than left to rot. BASIS OF THE CONSTANTS (worker-sized, 2026-07-31, from run 30602520347 @ 0ecb898, host srv1-01 arm64, post-#7490 fixes): whole-tree pass 258s over the ~1648-unit whole-tree entry closure (units = the leg's compiled closure module count, receipted by the first [compile-clean-cost] execution — NOT the ~2725-file module index, which overcounts by pool files outside every entry closure) = ~157ms/unit observed; per_unit_ms 320 is ~2x that rate, funding a full 2x fleet-spread host on top of the observed wall (the family margin ruling: a breach means content grew, never which host answered; pre-fix completing runs spread ~1.8x across hosts) — the batch family's 1000ms coefficient carries 1.4-1.7x the same way; overhead 60s covers the closure-size-independent tail (census fill parse ~13s fleet + governor arm). Whole-tree clamp = 60s + 1648*320ms = ~587s (~2.3x the observed 258s fleet wall). MECHANISM RECEIPTS (local x86, logic-not-cost host): green e2e via gunbc_ci_plan_artifact_plan — pass 310127ms/1648 units, WithinBudget, 1204 module rows, drift basis_absent=1205 counted, exit 0 (run at the pre-resize 200ms sizing; the arm proof is arithmetic-independent and the formula is witness-pinned at the final constants); RED control via GUNBC_FLOOR_COMPILE_CLEAN_BUDGET_TIGHTEN_MS=1 receipted beside it. DISPATCH: operator direction this session ('can we add the budget gating ... i suggest we share that model' — briansrls, 2026-07-31). OPERATOR SIGNATURE (briansrls, 2026-07-31): the worker-sized constants are affirmed as proposed — overhead_seconds 60, per_unit_ms 320 ('the clamp constants you chose are fine, please proceed') — so the refusal arm is signed-live per the family raise discipline; TIGHTENING may still land by ordinary receipt note, and the first fleet observations should tighten per_unit_ms toward the measured rate."
 
 data gunbc_ci_compile_clean_clamp: RunnableBatchClamp = RunnableBatchClamp {

dag/gunbc/explicit_witness_admission.dag

@@ -172,6 +172,14 @@ data explicit_witness_admissions: List<ExplicitWitnessAdmission> = [
     reason: "the english-language emit-ingest round trip returns false undiagnosed. Owner: the S2 emit family / behavioral-closure lane.",
     dissolution: unbound_dissolution(description: "the S2 behavioral-closure lane greens it locally, then this row deletes")
   ),
+  known_red_probe(
+    entry: "src/v1/tests/claim/occurrence_binding_parser_walk_witness_test.dag",
+    f: "parser_sibling_match_arm_pattern_binder_holds",
+    kind: CorpusWitnessKind,
+    budget: FastLaneEvalBudget,
+    reason: "STRUCTURALLY RED FROM AUTHORED SOURCE, and the parser is the reason rather than the witness. v1.compiler.parse stamps a match-arm pattern binder with the ARM's ancestors, so the binder is a SIBLING of the arm body rather than an ancestor of it; the binder's containment-derived LexicalExposure therefore has an exposing scope that does not contain the arm-body reference, and the isolation this witness asserts -- each arm's reference binding its own binder and refusing the other arm's -- cannot hold from authored source however the exposure authority is written. The file's own parser_sibling_match_arm_pattern_binder_holds_note has said so since the B1 rollup; what it lacked was an admission row, so the function stayed an ordinary green-required test that happened to be selected only when something in its import closure changed. It was newly selected by gunbc#8115's dag/std/occurrence_binding_resolve change and read as that PR's regression; it is not, and two bisect cycles on the wrong subject were the cost of the missing row. The equivalent property IS discharged, on a hand-built transport where the arm body nests under the binder occurrence, by occurrence_binding_candidates_witness -- so this row loses no coverage, it stops a known-structural red from being rediscovered as a new one.",
+    dissolution: unbound_dissolution(description: "the parser nests arm-body references under pattern-binder occurrences, or an equivalent exposure derivation lands that does not require that nesting; then this witness greens from authored source, this row deletes, and the function re-enrolls in namespace_reference_derived_closure_parser_integration_holds as its own note directs")
+  ),
   known_red_probe(
     entry: "dag/test/claim/legacy_test_behavior_disposition_acceptance_test.dag",
     f: "legacy_test_behavior_unclassified_frontier_is_zero",