P-037 Phase B / B1-F2-F4-R2: name T_B/R_B at the corrected instrument

named_later.T_B moves to e4199ec (the
B1-F2-F4-R2 head: 11f08d5's dump.rs carve-out plus the generic carve-out/
mutable-item invariant), named only now that both its own CI and the
evidence-only retake commit's CI (52a55c8)
are independently confirmed terminal-green with fast Kani 15/15 from the
raw logs -- never before either was itself proven green, matching every
prior naming commit's own discipline.

The prior named T_B (718a673, honest at the instrument boundary
B1-F2-F4-R1 held) is preserved as historical in a new
superseded_by_b1_f2_f4_r2 block, alongside superseded_by_b1_f2_f4_r1/
superseded_by_b1_f2_f2/superseded_by_b1_f2 -- never rewritten, only
superseded, exactly as every prior instrument correction in this epoch
has been recorded.

B_treatment_head and B_after_evidence stay null: no semantic treatment
exists yet. Naming-only: touches exactly docs/evidence/p037-b-epoch.json.
Both diff gates report IDENTICAL against 3ac13fd; ruff, mypy, and the
full test suite are clean at this commit.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AN6xHbpovxrmZ4AMjS7WQA

Author: claude

docs/evidence/p037-b-epoch.json

@@ -329,12 +329,19 @@
     "R_B": "docs/evidence/p037-b-baseline-manifest.md",
     "status": "historical: honest B1-F2-F2 evidence at an instrument boundary later corrected twice (B1-F2-F4, B1-F2-F4-R1); not a baseline for any future comparison; superseded by the T_B/R_B this same field family names once B1-F2-F4-R1's own retake lands (see named_later below)"
   },
-  "named_later": {
+  "superseded_by_b1_f2_f4_r2": {
+    "reason": "B1-F2-F4-R2 found that rust/crates/own-bridge/src/dump.rs -- already authorized by production_diff_gate.rust as a required B2.1b/c treatment surface (fn dump_summaries, item-level, since B1-F2-F4) -- had never been added to scripts/p037_evidence_b.py's own INSTRUMENT_CARVE_OUTS, so the already-authorized treatment would itself have moved instrument_identity between R_B and B_after and been refused as ineligible by comparison_problems() no matter how correct the guarded-summary solver was (11f08d5). A follow-up commit then closed the underlying CLASS of that defect, not just the instance: _record_closure_problems() now derives the expected Rust treatment carve-out set directly from production_diff_gate.rust's own mutable_items/mutable_files/unit (excluding controls) plus the extractor's own whole-unit carve-out, and requires exact-set equality against INSTRUMENT_CARVE_OUTS -- catching both a missing and an unjustified-excess carve-out, proven by a hostile selftest against a deep copy of the real committed epoch record (e4199ec). Both tasks changed the Phase-B measurement instrument (scripts/p037_evidence_b.py, a tracked INSTRUMENT_PATHS file) and its governance record, never production semantics -- both production diff gates report IDENTICAL between 3ac13fd4d41fcc8840d49005663146fd9188b8b4 (the closed B1-F2-F4-R1 naming head) and e4199ec4aa8d8f95c1b7e881f1ce3ba85f6ba58c (this retake's own T_B), reconfirmed directly, not assumed. The measurements below were honest at the instrument boundary B1-F2-F4/B1-F2-F4-R1 held -- that boundary was incomplete, not the measurement -- so they are superseded, never deleted or rewritten.",
     "T_B": "718a673c01641883df433b89cb76d5e3942d8815",
     "T_B_title": "P-037 Phase B / B1-F2-F4-R1: fix three instrument defects found by review",
     "R_B": "docs/evidence/p037-b-baseline-manifest.md",
+    "status": "historical: honest B1-F2-F4-R1 evidence at an instrument boundary later found incomplete (the dump.rs carve-out omission and the generic class it belonged to); not a baseline for any future comparison; superseded by the T_B/R_B this same field family names once B1-F2-F4-R2's own retake lands (see named_later below)"
+  },
+  "named_later": {
+    "T_B": "e4199ec4aa8d8f95c1b7e881f1ce3ba85f6ba58c",
+    "T_B_title": "P-037 Phase B / B1-F2-F4-R2: close the carve-out invariant generically",
+    "R_B": "docs/evidence/p037-b-baseline-manifest.md",
     "B_treatment_head": null,
     "B_after_evidence": null,
-    "note": "filled by B1-F2-F4-R1, naming last per the same order used every prior time (the corrected instrument head, 718a673, was already terminal-green -- CI 32/32, fast Kani 15/15 independently reconfirmed from the raw log, per B1-F2-F4-R1's own final report -- THEN four fresh, sequential, individually-verified M3 R_B takes were measured against it and copied into docs/evidence/ -- THEN that evidence-only commit (9d914e33f67f922da2218dd3e276252fa83a3638) went itself terminal-green -- CI 32/32, fast Kani 15/15 independently reconfirmed from the raw log -- THEN this naming commit). B_treatment_head/B_after_evidence stay null: no semantic treatment exists yet -- B2.1a is separate, later, and not yet authorized. See superseded_by_b1_f2_f4_r1 for what T_B/R_B named before this retake, superseded_by_b1_f2_f2 for the one before that, and superseded_by_b1_f2 for the one before that."
+    "note": "filled by B1-F2-F4-R2, naming last per the same order used every prior time (the corrected instrument head, e4199ec, was already terminal-green -- CI 32/32, fast Kani 15/15 independently reconfirmed from the raw log, per both B1-F2-F4-R2 commits' own CI runs -- THEN four fresh, sequential, individually-verified M3 R_B takes were measured against it and copied into docs/evidence/ -- THEN that evidence-only commit (52a55c8964f7aba0da0e7c93ac4f1475d7984283) went itself terminal-green -- CI 32/32, fast Kani 15/15 independently reconfirmed from the raw log -- THEN this naming commit). B_treatment_head/B_after_evidence stay null: no semantic treatment exists yet -- B2.1a is separate, later, and not yet authorized. See superseded_by_b1_f2_f4_r2 for what T_B/R_B named before this retake, superseded_by_b1_f2_f4_r1 for the one before that, superseded_by_b1_f2_f2 for the one before that, and superseded_by_b1_f2 for the one before that."
   }
 }