docs(B-0164.1): clarify remaining disagreement wiring

docs/backlog/P1/B-0164-dual-loop-substrate-attribution-and-reconciliation-protocol-2026-05-02.md

@@ -4,7 +4,7 @@ priority: P1
 status: open
 title: Dual-loop substrate attribution + reconciliation protocol — implementation work for BFT-many-masters at loop layer (Aaron 2026-05-02 + Otto independent extension)
 created: 2026-05-02
-last_updated: 2026-05-10
+last_updated: 2026-05-31
 depends_on:
   - B-0160
 children:
@@ -50,7 +50,7 @@ Aaron's morning reconciliation has to resolve these. The architecture's commitme
 
 1. **Per-loop attribution channel.** ✅ SATISFIED BY EXISTING SUBSTRATE — Otto's tick-shards under `docs/hygiene-history/ticks/**` already include col2 model-identifier (e.g., `opus-4-7 / autonomous-loop session continuation`). Codex's loop should write to the same shard surface with its own model-identifier (e.g., `gpt-5.5 / codex-loop`). The col2 schema accommodates this; no schema change needed.
 
-2. **Disagreement-preservation protocol for PR reviews.** ⏳ BLOCKED — requires dual-loop running. When both loops review the same PR thread with different conclusions: each loop's review-comment is preserved with attribution; neither auto-resolves the other. The morning reconciliation reads both, decides, resolves accordingly. No silent overwrites.
+2. **Disagreement-preservation protocol for PR reviews.** ⏳ PARTIAL — B-0164.1 has landed the pure detector/writer, divergence-shard builder, pending-shard reader, and reconcile CLI. Remaining work: wire a live PR-review workflow path so two loop observations on the same GitHub review thread actually invoke `fileReviewThreadDisagreement` when their conclusions differ.
 
 3. **Branch-attribution for in-flight work.** ✅ SATISFIED BY EXISTING SUBSTRATE — When both loops produce concurrent commits to the same branch: each loop commits with its own author-identifier (already supported by `Co-Authored-By` trailer). Different loops working different threads don't conflict if file-isolation holds; if they touch the same file, the second loop's tick should detect the conflict and either rebase OR file a divergence-shard noting the conflict for morning reconciliation.
 
@@ -134,14 +134,14 @@ ACs #1, #3, #4 are closed. Remaining open work extracted as atomic child rows:
 
 | Child | AC | Blocker | File |
 |-------|----|---------|------|
-| B-0164.1 | AC #2 — PR-review disagreement-preservation | B-0160 (dual-loop not running) | `B-0164.1-pr-review-disagreement-preservation-protocol.md` |
+| B-0164.1 | AC #2 — PR-review disagreement-preservation | Live PR-review caller wiring | `B-0164.1-pr-review-disagreement-preservation-protocol.md` |
 | B-0164.2 | AC #5 — Multi-loop tick-tooling attribution | B-0163 (tooling retirement) | `B-0164.2-multi-loop-tick-tooling-attribution.md` |
-| B-0164.3 | AC #6 — Cron-tick coordination | B-0160 + B-0164.1 | `B-0164.3-cron-tick-coordination-dual-loop.md` |
+| B-0164.3 | AC #6 — Cron-tick coordination | B-0164.1 live caller + topology observation | `B-0164.3-cron-tick-coordination-dual-loop.md` |
 
 **Dependency order:**
 
 ```
-B-0160  ──→  B-0164.1  ──→  B-0164.3
+B-0160 (closed) ──→  B-0164.1 caller wiring  ──→  B-0164.3
 B-0163  ──→  B-0164.2  ──→  (B-0164.3 write-safety audit)
 ```
 

docs/backlog/P1/B-0164.1-pr-review-disagreement-preservation-protocol.md

@@ -9,7 +9,7 @@ last_updated: 2026-05-31
 depends_on:
   - B-0160
 parent: B-0164
-classification: blocked
+classification: buildable-now
 decomposition: decomposed
 owners: [architect]
 type: friction-reducer
@@ -66,12 +66,18 @@ glue needed to detect that two loops have reviewed the same thread.
 existing `gh pr review` / `gh pr comment` surface); changing the divergence
 shard schema (already landed in B-0164 AC #4).
 
-## Blocker
+## Current blocker
 
-Requires two loops actually running concurrently (B-0160 harness side).
-Cannot test or verify the protocol without at least a simulated dual-loop
-execution. Unblock path: land B-0160 → stand up Codex loop → run a
-controlled dual-review experiment → validate this protocol against the outcome.
+The original B-0160 harness-side dependency is now closed, and the repo-native
+detector/reader/writer pieces have landed. The remaining work is the live
+caller: a PR-review workflow path must capture two loop observations for the
+same GitHub review thread and invoke `fileReviewThreadDisagreement` when their
+machine-comparable conclusions differ.
+
+Until that caller exists, AC #2 is not fully satisfied: a real dual-loop PR
+review can still produce differing conclusions without automatically filing a
+divergence shard. B-0164.3 remains separate; it decides tick cadence/topology
+after this row has an operational caller.
 
 ## Composes with
 
@@ -160,3 +166,27 @@ controlled dual-review experiment → validate this protocol against the outcome
   keeps list mode separate from the write-back path.
 - Focused tests cover argv parsing, usage-error no-write behavior, and an
   end-to-end CLI write-back against a writer-emitted divergence shard.
+
+## Current implementation state (2026-05-31)
+
+Implemented slices:
+
+- `detectReviewThreadDisagreement` and `fileReviewThreadDisagreement` in
+  `tools/hygiene/divergence-shard.ts` compare two loop observations for the
+  same PR review thread, avoid GitHub auto-resolution, and file a divergence
+  shard when machine-comparable conclusions differ.
+- `buildDivergenceShard` records the PR/thread topic, both attributed loop
+  perspectives, and a neutral disagreement summary, making the shard readable
+  without external context.
+- `scanDivergenceDir` plus `tools/hygiene/divergence-reconcile.ts --list` /
+  `--reconcile` provide the morning read/action path for pending shards.
+
+Remaining slice:
+
+- Wire an existing PR-review workflow or loop observation path to call
+  `fileReviewThreadDisagreement` after two loops review the same GitHub review
+  thread. A repo-wide search currently finds the detector/writer only in the
+  hygiene module and tests, so no live path files the shard yet.
+
+B-0164.3 remains open for the dual-loop cron-topology decision and observation
+window after this caller exists.

docs/backlog/P1/B-0164.3-cron-tick-coordination-dual-loop.md

@@ -5,9 +5,8 @@ status: open
 title: "Cron-tick coordination for dual-loop operation (dual-loop AC #6)"
 effort: S
 created: 2026-05-10
-last_updated: 2026-05-10
+last_updated: 2026-05-31
 depends_on:
-  - B-0160
   - B-0164.1
 parent: B-0164
 classification: blocked
@@ -75,13 +74,14 @@ The trade-off maps to what failure mode is higher-priority:
 
 ## Blocker
 
-Requires B-0160 (dual-loop harness side) and B-0164.1 (PR-review disagreement
-protocol — coordination choice affects review ordering). Cannot implement
-without at least two loops running experimentally.
+B-0160 is closed, so the harness-side prerequisite is no longer the active
+blocker. This row now waits on B-0164.1's live PR-review caller wiring: the
+cron topology decision should observe a system where differing loop conclusions
+on the same PR review thread can actually file a divergence shard.
 
 ## Composes with
 
-- B-0160 (dual-loop harness integration — prerequisite)
+- B-0160 (dual-loop harness integration — closed prerequisite)
 - B-0164 (parent — divergence-shard schema)
 - B-0164.1 (PR-review disagreement protocol — affected by topology choice)
 - B-0164.2 (tick-tooling attribution — concurrent writes need write-safety audit)