Bug type
Behavior bug (incorrect output/state without crash)
Beta release blocker
No
Summary
Gateway reset can rotate the OpenClaw session while preserving enough native Codex app-server state to resume the old Codex thread after reset.
For the Codex harness, reset needs to clear the native app-server sidecar binding and avoid carrying old native resume handles into the new session.
Parent tracker: #66251
Steps to reproduce
- Configure an agent to use the native Codex harness.
- Run a Codex-backed turn so the app-server creates a native thread binding.
- Reset the session through the gateway reset path.
- Send another Codex-backed turn after reset.
- Observe whether Codex resumes the old app-server thread instead of starting a fresh one.
Expected behavior
After gateway reset:
- the old Codex app-server binding is cleared
- the new OpenClaw session does not reuse stale native Codex resume state
- the next Codex turn starts a fresh native Codex thread unless the user explicitly requested resume behavior
Actual behavior
The gateway reset path can preserve or reuse the session file/native binding state that the Codex app-server uses to decide whether to resume a thread.
That can cause the post-reset session to continue with old Codex native context, even though the OpenClaw session has been reset.
OpenClaw version
2026.4.10 / 2026.4.11 reports, plus current main investigation
Operating system
macOS and Linux reports
Install method
npm/global CLI and local source investigation
Model
codex/gpt-5.4
Provider / routing chain
OpenClaw -> bundled codex plugin -> Codex app-server
Additional provider/model setup details
NOT_ENOUGH_INFO
Logs, screenshots, and evidence
Scope boundary
This issue is specifically about the gateway reset path and native Codex app-server bindings.
It should not track unrelated transcript replay, generic session reset behavior, or non-Codex CLI resume behavior unless it directly affects Codex app-server reset correctness.
Acceptance criteria
- Gateway reset invokes registered agent harness reset hooks for the old session.
- Codex reset clears the app-server binding associated with the old session file.
- Generated session reset gets a fresh generated transcript/session file where appropriate.
- Native Codex resume handles are not carried into the post-reset session accidentally.
- The next Codex turn after reset starts a new app-server thread rather than resuming the old one.
- Behavior is covered for gateway reset entrypoints, not only chat-command reset paths.
Suggested test coverage
- Gateway reset calls the registered Codex harness reset hook.
- Gateway reset removes or ignores the old Codex app-server sidecar binding.
- A post-reset Codex run calls native
thread/start instead of native thread/resume.
- Existing behavior for explicit resume/custom session file cases remains intentional.
Related
Impact and severity
Affected: Users running OpenClaw with the bundled Codex harness.
Severity: High for reset/new-session isolation in Codex-backed gateway sessions.
Frequency: Observed during Codex harness stabilization investigation.
Consequence: A reset session can accidentally retain old native Codex context, making reset behavior misleading and potentially carrying stale state forward.
Additional information
The local investigation has in-progress changes in this area. This issue is intended to capture the observed failure mode and the acceptance criteria for a focused fix.
Bug type
Behavior bug (incorrect output/state without crash)
Beta release blocker
No
Summary
Gateway reset can rotate the OpenClaw session while preserving enough native Codex app-server state to resume the old Codex thread after reset.
For the Codex harness, reset needs to clear the native app-server sidecar binding and avoid carrying old native resume handles into the new session.
Parent tracker: #66251
Steps to reproduce
Expected behavior
After gateway reset:
Actual behavior
The gateway reset path can preserve or reuse the session file/native binding state that the Codex app-server uses to decide whether to resume a thread.
That can cause the post-reset session to continue with old Codex native context, even though the OpenClaw session has been reset.
OpenClaw version
2026.4.10 / 2026.4.11 reports, plus current
maininvestigationOperating system
macOS and Linux reports
Install method
npm/global CLI and local source investigation
Model
codex/gpt-5.4
Provider / routing chain
OpenClaw -> bundled codex plugin -> Codex app-server
Additional provider/model setup details
NOT_ENOUGH_INFO
Logs, screenshots, and evidence
Scope boundary
This issue is specifically about the gateway reset path and native Codex app-server bindings.
It should not track unrelated transcript replay, generic session reset behavior, or non-Codex CLI resume behavior unless it directly affects Codex app-server reset correctness.
Acceptance criteria
Suggested test coverage
thread/startinstead of nativethread/resume.Related
Impact and severity
Affected: Users running OpenClaw with the bundled Codex harness.
Severity: High for reset/new-session isolation in Codex-backed gateway sessions.
Frequency: Observed during Codex harness stabilization investigation.
Consequence: A reset session can accidentally retain old native Codex context, making reset behavior misleading and potentially carrying stale state forward.
Additional information
The local investigation has in-progress changes in this area. This issue is intended to capture the observed failure mode and the acceptance criteria for a focused fix.