Best possible solution:

Land the narrow queued fallback after real Telegram/Codex proof confirms exactly-once deferred delivery, while keeping same-turn transcript-backed steering disabled until Codex exposes durable idempotent acceptance.

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

Yes at source level and in the targeted regression setup: current main can reject transcript-backed steering and continue without deferring the durable inbound message. No live Telegram/Codex artifact demonstrates the current or fixed behavior.

Is this the best way to solve the issue?

Yes. The patch uses the established queued lifecycle and is safer than best-effort steering because current Codex appends each accepted steer and does not deduplicate clientUserMessageId retries.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against 1bcc4c5e7063.

Label changes:

• add P1: The bug can suppress a durable Telegram message during a real active Codex turn.
• add merge-risk: 🚨 message-delivery: The changed admission branch determines whether a rejected steer is preserved for later delivery.
• add merge-risk: 🚨 session-state: Correctness relies on transferring ownership to the queued lifecycle and delaying durable adoption until later reply-lane admission.
• add rating: 🦪 silver shellfish: Overall readiness is 🦪 silver shellfish; proof is 🦪 silver shellfish and patch quality is 🐚 platinum hermit.
• add status: 📣 needs proof: The PR needs real behavior proof before ClawSweeper can clear the contributor ask. Needs real behavior proof before merge: The PR provides only mocked end-to-end regression coverage; add redacted live Telegram/Codex logs, output, or a recording, update the PR body, and if review does not refresh ask a maintainer to comment @clawsweeper re-review.
• add mantis: telegram-visible-proof: Mantis should capture Telegram visible proof. The changed behavior is directly demonstrable as a Telegram follow-up reply instead of a lost message.

Label justifications:

• P1: The bug can suppress a durable Telegram message during a real active Codex turn.
• merge-risk: 🚨 message-delivery: The changed admission branch determines whether a rejected steer is preserved for later delivery.
• merge-risk: 🚨 session-state: Correctness relies on transferring ownership to the queued lifecycle and delaying durable adoption until later reply-lane admission.
• rating: 🦪 silver shellfish: Overall readiness is 🦪 silver shellfish; proof is 🦪 silver shellfish and patch quality is 🐚 platinum hermit.
• status: 📣 needs proof: The PR needs real behavior proof before ClawSweeper can clear the contributor ask. Needs real behavior proof before merge: The PR provides only mocked end-to-end regression coverage; add redacted live Telegram/Codex logs, output, or a recording, update the PR body, and if review does not refresh ask a maintainer to comment @clawsweeper re-review.
• mantis: telegram-visible-proof: Mantis should capture Telegram visible proof. The changed behavior is directly demonstrable as a Telegram follow-up reply instead of a lost message.

PR surface:

Source +4, Tests +29. Total +33 across 2 files.

What I checked:

• Current-main failure path: Current main requests transcript commit waiting when durable adoption is required; after a rejected steer, steer-mode admission can continue without forcing the inbound message into the follow-up queue. (src/auto-reply/reply/agent-runner.ts:1286, 1bcc4c5e7063)
• Proposed fallback is narrowly scoped: The PR forces follow-up admission only for transcript_commit_wait_unsupported; other steering rejection reasons retain their existing behavior. (src/auto-reply/reply/agent-runner.ts:1320, 348182ab12a8)
• Telegram deferred-adoption lifecycle: Telegram marks a queued turn deferred at enqueue, performs durable adoption when the queued turn is admitted, and abandons it if queue ownership completes without admission. (extensions/telegram/src/bot-message-dispatch.ts:2651, 1bcc4c5e7063)
• Follow-up admission boundary: The follow-up runner calls the queued lifecycle admission hook only after acquiring reply-lane ownership, so durable ingress adoption precedes later-turn execution. (src/auto-reply/reply/followup-runner.ts:720, 1bcc4c5e7063)
• Codex dependency contract: Codex appends every accepted steer as a new pending UserInput; clientUserMessageId is carried as correlation metadata and is not checked as an idempotency key. (../codex/codex-rs/core/src/session/mod.rs:3931, a328f30172c6)
• Upstream capability remains missing: The open Codex issue requests durable retry idempotency for turn/steer, confirming that OpenClaw cannot safely treat current RPC acceptance as authoritative transcript adoption.

Likely related people:

• obviyus: Ayaan Zaidi introduced the channel turn-adoption acknowledgement and recently maintained active-run steering gates. (role: introduced behavior; confidence: high; commits: ef103205adab, c2789b52a7fd; files: src/auto-reply/reply/agent-runner.ts, src/agents/embedded-agent-runner/runs.ts)
• vincentkoc: Vincent Koc authored and merged the current Telegram deferred, admitted, and abandoned spooled-turn lifecycle that this fallback uses. (role: recent lifecycle owner; confidence: high; commits: f79e8859855b; files: extensions/telegram/src/bot-message-dispatch.ts, src/auto-reply/reply/queue/types.ts, src/auto-reply/reply/agent-runner.ts)
• steipete: Peter Steinberger recently maintained admitted follow-up ownership and restart behavior on current main, which shares the deferred-work invariant. (role: adjacent recent contributor; confidence: medium; commits: 1bcc4c5e7063; files: src/auto-reply/reply/agent-runner.ts, src/auto-reply/reply/followup-runner.ts)

• 🦀 challenger crab: rare, exceptional readiness with strong proof, clean implementation, and convincing validation.
• 🦞 diamond lobster: very strong readiness with only minor maintainer review expected.
• 🐚 platinum hermit: good normal PR, likely mergeable with ordinary maintainer review.
• 🦐 gold shrimp: useful signal, but proof or patch confidence is still limited.
• 🦪 silver shellfish: thin signal; proof, validation, or implementation needs work.
• 🧂 unranked krab: not merge-ready because proof is missing/unusable or there are serious correctness or safety concerns.
• 🌊 off-meta tidepool: rating does not apply to this item.

Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics.

• ClawSweeper keeps one durable marker-backed review comment per issue or PR.
• Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
• A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
• PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
• Maintainers can also comment @clawsweeper review to request a fresh review only.
• Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
• Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
• Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.