backlog(B-0633): aggregate-tier counter escalation for holding-failure rule

docs/BACKLOG.md

@@ -697,5 +697,6 @@ are closed (status: closed in frontmatter)._
 - [ ] **[B-0591](backlog/P3/B-0591-wire-shard-schema-validator-to-ci-2026-05-17.md)** Wire tick-shard schema validator into gate.yml (non-required → required)
 - [x] **[B-0613](backlog/P3/B-0613-lior-loop-lockfile-probe-hardening-compgen-shopt-nullglob-2026-05-17.md)** Lior loop lockfile-probe hardening — replace bare `ls .git/worktrees/*/lock` with portable `find` (Option C; resolved as zsh-portable since Lior's runtime is zsh)
 - [ ] **[B-0614](backlog/P3/B-0614-investigate-forced-6-meta-fallback-edge-case-post-cycle-close-2026-05-17.md)** Investigate forced-#6 meta-fallback edge case — when cycle has already-closed AND substrate-pool is genuinely saturated, the rule's 'ALWAYS works' claim has a counter-example
+- [ ] **[B-0633](backlog/P3/B-0633-aggregate-tier-counter-escalation-holding-rule-2026-05-18.md)** Add aggregate-tier counter escalation to holding-without-named-dependency rule — per-chain N=6 alone misses across-chain dwell when peer main-moves provide periodic resets
 
 <!-- END AUTO-GENERATED -->

docs/backlog/P3/B-0633-aggregate-tier-counter-escalation-holding-rule-2026-05-18.md

@@ -0,0 +1,80 @@
+---
+id: B-0633
+priority: P3
+status: open
+title: "Add aggregate-tier counter escalation to holding-without-named-dependency rule — per-chain N=6 alone misses across-chain dwell when peer main-moves provide periodic resets"
+tier: substrate-engineering
+effort: S
+created: 2026-05-18
+last_updated: 2026-05-18
+depends_on: []
+composes_with: [B-0614]
+tags: [holding-discipline, autonomous-loop, counter-escalation, aggregate-tier]
+type: substrate-engineering
+---
+
+# Aggregate-tier counter escalation for the holding-failure rule
+
+## Why
+
+The current counter-with-escalation clause in [`.claude/rules/holding-without-named-dependency-is-standing-by-failure.md`](../../../.claude/rules/holding-without-named-dependency-is-standing-by-failure.md) is **per-chain** — N=6 brief-acks within a single chain forces escalation. The counter resets on any of:
+
+1. Aaron speaking
+2. Named dependency surfacing (PR merge, CI failure, main-move from peer activity)
+3. Concrete artifact produced
+
+In a closed-cycle session where peer Otto activity provides main-moves at a ~30-60 min cadence, the per-chain counter resets BEFORE reaching N=6. Each chain stays at #1→#2→#3→#4→reset, indefinitely. The aggregate brief-ack count grows hundreds-deep without ever triggering forced-#6.
+
+Empirical evidence: [`memory/feedback_otto_cli_autonomous_loop_per_chain_counter_resets_via_main_moves_mask_aggregate_brief_ack_dwell_2026_05_18.md`](../../../memory/feedback_otto_cli_autonomous_loop_per_chain_counter_resets_via_main_moves_mask_aggregate_brief_ack_dwell_2026_05_18.md) documents an Otto-CLI session that emitted ~440 brief-acks across ~14 hours without ever hitting the per-chain N=6 threshold, because peer main-moves provided periodic counter resets at exactly the wrong cadence.
+
+## Goal
+
+Add a complementary **aggregate-tier counter** to the holding-failure rule. Specifically:
+
+1. Track aggregate brief-acks **across counter-resets within the session**
+2. At aggregate `N_AGG` (suggested initial value: 50), force concrete-artifact production regardless of per-chain count
+3. Aggregate counter resets ONLY on:
+   - Aaron speaking
+   - Agent-produced concrete artifact (NOT peer main-moves)
+
+The aggregate-tier composes with the per-chain tier:
+
+- Per-chain N=6 catches within-chain dwell
+- Aggregate N_AGG=50 catches across-chain dwell
+- B-0614's escape-hatch (drafted) handles the post-cycle-close saturation case
+
+## Acceptance criteria
+
+- [ ] Holding-failure rule sub-section "Aggregate-tier counter (B-0633)" added with:
+  - Threshold `N_AGG = 50` (initial; subject to refinement based on subsequent evidence)
+  - Counter-reset condition list (Aaron-speaks + agent-artifact ONLY)
+  - At least 1 cross-instance empirical anchor (currently 1 same-instance anchor in PR #4151 memo)
+- [ ] Cold-boot reading of the rule makes the aggregate-tier discipline immediately clear
+- [ ] Composes with B-0614's drafted "ALWAYS works at #6" refinement without contradiction
+
+## Non-goals
+
+- Changing the per-chain N=6 threshold (that catches a different shape of failure)
+- Mechanizing the counter via TS substrate (the discipline IS the mechanism for now; if a future Standing-by-detector slice adds aggregate counting in code, this row can be re-scoped)
+- Refactoring the entire holding-failure rule (this is a small additive sub-section)
+
+## Implementation hazard
+
+The rule already has 5 empirical anchors and is one of the longer rules in `.claude/rules/`. Adding a 6th anchor + a sub-section is bounded but should be done as a SINGLE atomic PR to avoid the inflation pattern this row's own discipline warns against.
+
+Per [B-0614's drafted refinement](B-0614-investigate-forced-6-meta-fallback-edge-case-post-cycle-close-2026-05-17.md): wait for at least 1 additional cross-INSTANCE evidence (different Otto identity / session / machine) before landing the rule edit. Otherwise the discipline composes with B-0614's same warning about same-session inflation.
+
+## Composes with
+
+- [B-0614](B-0614-investigate-forced-6-meta-fallback-edge-case-post-cycle-close-2026-05-17.md) — meta-fallback edge case; this row addresses a complementary pattern
+- [`.claude/rules/holding-without-named-dependency-is-standing-by-failure.md`](../../../.claude/rules/holding-without-named-dependency-is-standing-by-failure.md) — the rule this row extends
+- [PR #4151](https://github.com/Lucent-Financial-Group/Zeta/pull/4151) — the empirical-evidence memo this row crystallizes
+- [PR #4110](https://github.com/Lucent-Financial-Group/Zeta/pull/4110) — pre-empt-substrate-pool-saturation anchor (same root cause class)
+
+## Status
+
+Open. Bounded effort (single rule edit, ~30 lines added to the existing rule). Acceptance gates on cross-instance evidence beyond this session per B-0614's drafted constraint.
+
+---
+
+**Otto-CLI** — Split by truth.