Summary

• Problem: PR [PARALLEL] feat(continuation): substrate-native chain-budget extraction (path-b) #354 had two confirmed-real P1 post-compaction delegate queue regressions: filtered drains stranded failed retries, and chain-state persistence failures still allowed queue ack.
• Why it matters: transient spawn or session-store write failures could leave continuation delegates stuck until restart, or ack a spawned shard while durable chain state stayed stale.
• What changed: fresh post-compaction drains now run a targeted pass followed by an unfiltered retry pass; chain-state persistence failures rethrow so queue recovery fails/retries; lifecycle released-delegate telemetry now counts accepted fresh drain deliveries.
• What did NOT change (scope boundary): no queue storage format, config, retry cap, or gateway startup recovery contract changes.

Change Type (select all)

• Bug fix
• Feature
• Refactor required for the fix
• Docs
• Security hardening
• Chore/infra

Scope (select all touched areas)

• Gateway / orchestration
• Skills / tool execution
• Auth / tokens
• Memory / storage
• Integrations
• API / contracts
• UI / DX
• CI/CD / infra

Linked Issue/PR

• Related [PARALLEL] feat(continuation): substrate-native chain-budget extraction (path-b) #354
• Addresses codex review comments r3144331033, r3144344309, and optional r3144344310 on [PARALLEL] feat(continuation): substrate-native chain-budget extraction (path-b) #354.
• This PR fixes a bug or regression

Root Cause (if applicable)

• Root cause: extraction moved post-compaction delegate delivery through the durable session-delivery queue, but dispatch only drained newly queued entry IDs and chain-state persistence errors were logged instead of propagated.
• Missing detection / guardrail: no regression covered failed queued delegate retry across later drain cycles or session-store write failure after an accepted post-compaction spawn.
• Contributing context (if known): pre-extraction inline dispatch counted accepted spawns directly and did not split enqueue/drain/ack across queue recovery.

Regression Test Plan (if applicable)

• Coverage level that should have caught this:
  • Unit test
  • Seam / integration test
  • End-to-end test
  • Existing coverage already sufficient
• Target test or file: src/auto-reply/reply/post-compaction-delegate-dispatch.test.ts
• Scenario the test should lock in: failed post-compaction queue entries remain pending through backoff and are picked up by a later unfiltered drain; session-store write failure increments retry metadata instead of acking; lifecycle release count comes from accepted fresh drain deliveries.
• Why this is the smallest reliable guardrail: the dispatch test exercises the actual queue drain and delivery seams without requiring gateway startup or external provider processes.
• Existing test that already covers this (if any): existing queue storage/recovery tests cover retry metadata, backoff, and retry-budget exhaustion generically.
• If no new test is added, why not: N/A

User-visible / Behavior Changes

Operators should see post-compaction delegate retries recover during later drain cycles instead of waiting for restart, and lifecycle release counts no longer include queued-but-not-accepted delegate deliveries.

Diagram (if applicable)

Before:
enqueue fresh IDs -> targeted drain only -> failed old entries remain filtered out

After:
enqueue fresh IDs -> targeted drain -> lifecycle accepted count -> unfiltered retry drain -> eligible old failures retried

Security Impact (required)

• New permissions/capabilities? No
• Secrets/tokens handling changed? No
• New/changed network calls? No
• Command/tool execution surface changed? No
• Data access scope changed? No
• If any Yes, explain risk + mitigation: N/A

Repro + Verification

Environment

• OS: Linux
• Runtime/container: Node 22+ workspace via pnpm
• Model/provider: N/A (queue/delivery seam tests use mocks)
• Integration/channel (if any): N/A
• Relevant config (redacted): default test continuation runtime config

Steps

1. Enqueue a post-compaction delegate and make the first spawn/persistence attempt fail.
2. Run a later drain cycle without entryIds after retry backoff elapses.
3. Observe the entry retry and ack only after successful delivery/persistence.

Expected

• Failed entries remain queued with retry metadata and later eligible unfiltered drains retry them.
• updateSessionStore failure causes queue failure/retry, not ack.
• Lifecycle releasedDelegates reflects accepted fresh deliveries.

Actual

• Fixed by this PR.

Evidence

• Failing test/log before + passing after
• Trace/log snippets
• Screenshot/recording
• Perf numbers (if relevant)

Human Verification (required)

• Verified scenarios: pnpm tsgo; pnpm check; pnpm test src/auto-reply/reply/post-compaction-delegate-dispatch.test.ts; pnpm test src/infra/session-delivery-queue-storage.ts src/infra/session-delivery-queue-recovery.ts; pnpm check:additional; pnpm build; pnpm check:changed --base 3eb87a7 --head HEAD.
• Edge cases checked: targeted drains still bypass fresh-ID backoff; unfiltered retry drains respect backoff; persistence failure records retry metadata; lifecycle counts accepted fresh deliveries only.
• What you did not verify: live gateway/provider spawning outside the mocked seam.

Review Conversations

• I replied to or resolved every bot review conversation I addressed in this PR.
• I left unresolved only the conversations that still need reviewer or maintainer judgment.

Compatibility / Migration

• Backward compatible? Yes
• Config/env changes? No
• Migration needed? No
• If yes, exact upgrade steps: N/A

Risks and Mitigations

• Risk: A retry after persistence failure can be at-least-once because the first spawn may already have been accepted.
  • Mitigation: durable queue now avoids acking stale chain state; retry cap and backoff still bound repeated attempts.