board: three fruits — bgz-tensor tests ungated, causal-edge Revision pools duplicates, Fisher-z clamps silently

.claude/board/FRUITS.md

@@ -4,7 +4,7 @@ Generated by `.claude/tools/fruits_index.py --write`. Do not hand-edit:
 add or change a `**Fruit:**` line on the ledger row and regenerate.
 CI fails when this file is stale.
 
-Open: 21. Closed: 0.
+Open: 24. Closed: 0.
 
 ## discovery torque
 
@@ -26,6 +26,8 @@ Open: 21. Closed: 0.
 
 | tier | row | status | ledger |
 |---|---|---|---|
+| High-signal | TD-NARS-REVISION-UNGUARDED-CAUSAL-EDGE-ISA-1 — `Opcode::Revision` pools a duplicated source; a measured site of TD-NARS-REVISION-UNGUARDED (2026-10-23) | OPEN. **Basis:** VERIFIED-IN-CODE + MEASURED. | `TECH_DEBT.md` |
+| Low-hanging | TD-BGZ-TENSOR-INTEGRATION-TESTS-UNGATED-1 (2026-10-23) — OPEN | OPEN | `TECH_DEBT.md` |
 | Low-hanging | TD-CE64-W-SLOT-DOC-SAYS-NO-ANCHOR — `edge.rs` still documents W = 0 as "no corpus anchor" (2026-10-23) — OPEN | OPEN. **Basis:** VERIFIED-IN-CODE (`edge.rs:944` on main `8f610a85`). | `TECH_DEBT.md` |
 | Low-hanging | TD-STATUS-BOARD-ROWS-MISSING-FOR-OCT-PLANS-1 (2026-10-23) — OPEN | — | `TECH_DEBT.md` |
 
@@ -40,6 +42,7 @@ Open: 21. Closed: 0.
 | High-signal | ISS-D-CGW-4-PROMOTION-GATE — the counterfactual verdict is computed but promotes nothing (2026-10-23) | OPEN. **Basis:** VERIFIED-IN-CODE (`crates/lance-graph-planner/src/pearl.rs:259-263` `counterfactual_verdict` calls `Reaction::verdict`; callers found only in `pearl.rs` tests `:747-815`; STATUS_BOARD row D-CGW-4 Queued). | `ISSUES.md` |
 | High-signal | ISS-DURABLEREF-THOUGHTCOMMITID — the two identity types four plans depend on do not exist (2026-10-20) | OPEN. **Basis:** VERIFIED-IN-CODE (`DurableRef` and `ThoughtCommitId` appear in `crates/` only in doc comments, `mork-pathmap-probe/src/rmap.rs:70` and `:690`). | `ISSUES.md` |
 | High-signal | ISS-BF16-POLYFILL-PER-ELEMENT-ALLOC — ndarray's BF16 16×16 polyfill allocates per output element (2026-10-20) | OPEN. **Basis:** VERIFIED-IN-CODE (`ndarray/src/simd_ops.rs`, `bf16_tile_gemm_16x16`) and MEASURED (D-RMAP-2 §5a: a 256 × 256 `A·A` takes about 19 ms, about 0.9 GMAC/s, on an AVX-512 host without AMX or `avx512_bf16`). | `ISSUES.md` |
+| Low-hanging | ISS-FISHERZ-ENCODE-SATURATES-SILENTLY — out-of-family cosines encode to the edge code with no flag (2026-10-23) | OPEN. **Basis:** VERIFIED-IN-CODE (`crates/bgz-tensor/src/fisher_z.rs:67`, `i8_val.clamp(-128.0, 127.0) as i8`, read at `main` `8f610a85`) + MEASURED. | `ISSUES.md` |
 | Low-hanging | ISS-CAUSAL-EDGE-CARRIES-SEVEN-PRE-EXISTING-CLIPPY-FINDINGS (2026-08-22) — OPEN | — | `ISSUES.md` |
 | Heavy | ISS-ORDER-AND-CODEBOOK-WITNESS-HAS-NO-CONTRACT-TYPE — seven open items wait on one witness nobody owns (2026-10-23) — OPEN | OPEN (reframed, see correction). **Basis:** VERIFIED-IN-CODE | `ISSUES.md` |
 | Heavy | ISS-ADMISSION-B-CALL-SITE-CLASSIFICATION-NOT-STARTED — Semijoin, EqU32Via and Gather are unclassified, and no owner is named (2026-10-23) — OPEN | OPEN (operator decision on ownership). **Basis:** VERIFIED-IN-CODE | `ISSUES.md` |

.claude/board/ISSUES.md

@@ -1,3 +1,26 @@
+## ISS-FISHERZ-ENCODE-SATURATES-SILENTLY — out-of-family cosines encode to the edge code with no flag (2026-10-23)
+
+**Status:** OPEN. **Basis:** VERIFIED-IN-CODE (`crates/bgz-tensor/src/fisher_z.rs:67`, `i8_val.clamp(-128.0, 127.0) as i8`, read at `main` `8f610a85`) + MEASURED.
+**Fruit:** Low-hanging
+
+- `FisherZTable::encode` clamps anything beyond the family's
+  `[z_min, z_max]` onto the edge code, and nothing reports it. An
+  out-of-family pair therefore reads as maximal in-family similarity. Below
+  `z_min` the value lands on −128, which `decode` maps one step past `z_min`.
+- **Receipt:** P-D3, `crates/bgz-tensor/tests/p_d3_ood_support_probe.rs`
+  (`entries/2026-10-22-probes-pd4-pe2-pd3.md`). The clamp condition holds on
+  0/120 in-range cosines and 403/403 saturating ones, and
+  `encode(1.0) == encode(0.99995)`.
+- This is H5 of the H1–H5 falsifiers (EXO-50, session_01G22yT6 /tanh-01). H5
+  was proposed and never added.
+- **What closes it:** a checked encode that reports out-of-range values (a
+  count, or an `Option`) beside the existing one, with no change to the code
+  points. Add a can-fire test (a value past `z_max` is reported) and a
+  stay-silent test (in-family values are not).
+- Note: until TD-BGZ-TENSOR-INTEGRATION-TESTS-UNGATED-1 is fixed, a test under
+  `bgz-tensor/tests/` does not run in CI. Put this one in the crate's lib
+  tests.
+
 ## ISS-ORDER-AND-CODEBOOK-WITNESS-HAS-NO-CONTRACT-TYPE — seven open items wait on one witness nobody owns (2026-10-23) — OPEN
 
 ADMISSION B (`entries/2026-10-20-six-decisions.md` §1, ruled in

.claude/board/TECH_DEBT.md

@@ -1,3 +1,47 @@
+## TD-NARS-REVISION-UNGUARDED-CAUSAL-EDGE-ISA-1 — `Opcode::Revision` pools a duplicated source; a measured site of TD-NARS-REVISION-UNGUARDED (2026-10-23)
+
+**Status:** OPEN. **Basis:** VERIFIED-IN-CODE + MEASURED.
+**Fruit:** High-signal
+
+- `crates/causal-edge/src/isa.rs:209` `truth_code` (and `:197` `truth`) runs
+  `Opcode::Revision` through `truth::revision` on four scalars. There is no
+  stamp, so nothing can see that the two operands share evidence.
+- **Receipt:** P-E2, `crates/causal-edge/tests/p_e2_stamp_gate_probe.rs`
+  (`entries/2026-10-22-probes-pd4-pe2-pd3.md`). Revising one source with
+  itself k times starting from c = 128 gives c = 170 / 204 / 227 at
+  k = 2 / 4 / 8, with posterior sd ratio 0.837 / 0.658 / 0.491.
+- The test-local ONA-shape gate already exists in that file and is
+  disable-verified. It checks stamp overlap first, and on overlap keeps the
+  stronger belief. With it, duplicates give Δc = 0 and an sd ratio of 1.000,
+  while disjoint sources still pool.
+- This is the same defect as TD-NARS-REVISION-UNGUARDED (2026-07-23, ndarray
+  `nars.rs` and planner `nars_engine.rs`). That row does not list this site.
+- **Open:** where the stamp lives (G5 in the entry). The guard belongs at the
+  belief store, not in the pure truth function.
+- **What closes it:** pick the stamp's home, then call `Revision` only through
+  a gate that refuses or chooses on overlap. Move P-E2's three tests onto the
+  production path; each must go red when its guard is removed.
+
+## TD-BGZ-TENSOR-INTEGRATION-TESTS-UNGATED-1 (2026-10-23) — OPEN
+
+CI runs `bgz-tensor` with `--lib` only (`.github/workflows/rust-test.yml:274`),
+so nothing under `crates/bgz-tensor/tests/` runs in CI. The crate is
+workspace-excluded, so no other step reaches it.
+
+On main `8f610a85` that directory holds 9 lab probes, 23 tests. Among them are
+P-S2 (`p_s2_compose_chain_probe.rs`, #1726), the evidence behind TORQUE row
+T-2026-10-23-1, and P-D3. Draft #1731 adds the EXO-129 probe in the same
+directory, so its PR checks would pass without running it.
+
+**Receipt:** `cargo test --manifest-path crates/bgz-tensor/Cargo.toml --tests`
+on main `8f610a85` with `CARGO_PROFILE_{DEV,TEST}_DEBUG=0 CARGO_INCREMENTAL=0`:
+207 lib tests plus all 9 integration files (3+4+2+3+2+2+2+2+3 = 23 tests), 0 failed.
+No test reads a file or an env var, so nothing blocks gating them.
+**Fix:** drop `--lib` from that step (or add a second `--tests` step). The
+step comment gives no reason for `--lib`.
+**Fruit:** Low-hanging
+**Status:** OPEN
+
 ## TD-CE64-W-SLOT-DOC-SAYS-NO-ANCHOR — `edge.rs` still documents W = 0 as "no corpus anchor" (2026-10-23) — OPEN
 
 `crates/causal-edge/src/edge.rs:944`: "Witness corpus root handle (6-bit,