Best possible solution:

Close this conflicting replacement PR and continue the fix review on #90750.

Do we have a high-confidence way to reproduce the issue?

Yes. Source inspection shows current main leaves tool/model markers untouched when the reply-run wrapper ends with clearRunActivity: false, and the newer canonical PR supplies before/after runtime logs showing the blocked_tool_call path disappearing.

Is this the best way to solve the issue?

No. This PR was a plausible early repair, but current main and the newer clean PR make the narrower owner-less reply-run cleanup the better maintainable path.

Security review:

Security review cleared: No concrete security or supply-chain concern was found in this superseded diff; it touches runtime diagnostics/session cleanup and tests, not dependencies, workflows, secrets, or install scripts.

AGENTS.md: found and applied where relevant.

What I checked:

• Repository policy read: Read the full root AGENTS.md and the gateway-scoped AGENTS.md; the review-depth and gateway hot-path guidance apply because the PR touches gateway/session recovery code. (AGENTS.md:1, 663fabbe30eb)
• This PR is no longer a clean landing target: Live GitHub status for this PR reports mergeStateStatus DIRTY, mergeable CONFLICTING, and failed dependency-guard/checks-node-agentic-agents checks at head e32fc45. (e32fc453ae5a)
• Current main already has newer recovery-owned cleanup plumbing: Current main applies mutating recovery outcomes through clearDiagnosticEmbeddedRunActivityForSession in the recovery coordinator, so the older recovery/reset cleanup branch no longer matches the current owner boundary cleanly. (src/logging/diagnostic-session-recovery-coordinator.ts:135, 663fabbe30eb)
• Remaining source gap is the owner-less reply-run path: Current main still ends reply-run diagnostic work with clearRunActivity: false, and markDiagnosticEmbeddedRunEnded leaves tool/model markers untouched in that opt-out path; this is the narrower path targeted by the newer PR. (src/auto-reply/reply/reply-run-registry.ts:353, 663fabbe30eb)
• Canonical replacement is viable: The newer PR at fix(diagnostics): evict orphaned tool/model activity on owner-less run end #90750 closes the same issue, is MERGEABLE/CLEAN, has green checks including Real behavior proof, and supplies before/after runtime diagnostic logs for the same blocked_tool_call leak. (83c3212755ca)
• Related PR search: The live PR search for the linked issue shows this PR, the closed source PR, and the newer open canonical PR; no other active landing path was found for the same central fix.

Likely related people:

• 849261680: Authored the newer open PR that closes the same issue with a narrower owner-less reply-run cleanup and stronger runtime proof. (role: canonical follow-up PR author; confidence: high; commits: 83c3212755ca; files: src/logging/diagnostic-run-activity.ts, src/auto-reply/reply/reply-run-registry.test.ts, src/infra/diagnostic-events.ts)
• iloveleon19: Current-main blame attributes the central markDiagnosticEmbeddedRunEnded and reply-run diagnostic wrapper lines to this recent commit. (role: recent area contributor; confidence: medium; commits: 4892bbc10f46; files: src/logging/diagnostic-run-activity.ts, src/auto-reply/reply/reply-run-registry.ts)
• Vincent Koc: The latest release commit in this checkout carries the current diagnostics, recovery coordinator, reset cleanup, and gateway reset service files. (role: current source provenance; confidence: medium; commits: 8c802aa68351; files: src/logging/diagnostic-run-activity.ts, src/logging/diagnostic-session-recovery-coordinator.ts, src/gateway/session-reset-service.ts)
• sandieman2: Recently changed reply-run session-lane behavior in the same area that controls follow-up admission and queued reply delivery. (role: adjacent reply/session-lane contributor; confidence: medium; commits: c67dc59b02b0; files: src/auto-reply/reply/reply-run-registry.ts)

Codex review notes: model internal, reasoning high; reviewed against 663fabbe30eb.

Share on X: post this hatch
Copy: My PR egg hatched a 🥚 common Cosmic Crabkin in ClawSweeper.
Hatchability:

• Merged PRs are hatchable.
• Open PRs are hatchable when they are status: 👀 ready for maintainer look, status: 🚀 automerge armed, or labeled clawsweeper:automerge.
• Closed unmerged PRs are hatchable only when one of those hatchable labels is still present in the durable record.

About:

• Eggs appear after real-behavior proof passes. They are collectible flavor only.
• Review momentum changes the shell state: follow-up work warms it, re-review makes it wobble, and a clean final review lets it hatch.
• The hatch is seeded from this repository and PR number, so the same PR keeps the same creature; the reviewed head SHA can only change safe visual details.
• Rarity is just collectible sparkle: 🥚 common, 🌱 uncommon, 💎 rare, ✨ glimmer, and 🌈 legendary.