OpenClaw Issue Digest: Codex Runtime Regressions and Gateway Stability
Open Issues
Recent activity in the OpenClaw repository reveals a concentration of high-severity regressions centered around the Codex runtime and general gateway stability. A significant number of reports indicate that the transition to the native Codex runtime is currently unstable, with multiple users reporting MODULE_NOT_FOUND errors and timeout-driven fallbacks.
Codex Runtime and Model Routing
Several issues (#81213, #81196, #81175, #81194) highlight a critical failure in the Codex runtime's module resolution. Users updating to 2026.5.12-beta versions report that the primary OpenAI/Codex route fails to load codex-native-task-runtime, forcing the system to fall back to Anthropic providers. Even when the module is present on disk, the runtime appears to resolve the path incorrectly, treating a .cjs file as a directory.
Furthermore, the Codex app-server is experiencing stability issues during high-context turns. Issue #81114 describes a "wall-clock" timeout that does not distinguish between a dead turn and one that is still progressing near the context window limit, leading to premature model fallbacks.
Gateway Performance and Stability
Severe event loop starvation has emerged as a recurring theme. Issue #81191 reports delays of up to 7 minutes, which the author attributes to blocking HTTP calls during the Telegram channel's startAccount phase. Similarly, issue #81172 reports that the memory_search tool can block the Node.js event loop for over 60 seconds, causing Discord gateway connections to close and agents to hang.
Security and Data Integrity
Two high-severity security concerns have been raised:
- Secret Exposure (#71211): The
exectool returns raw stdout/stderr to the agent without redaction, potentially leaking API keys or.envcontents into the agent's context. - Auth Bypass (#72418): A vulnerability in
shouldSkipLocalBackendSelfPairingallows local processes with a valid gateway token to impersonate the internal backend client and bypass device pairing.
Additionally, a data-loss class bug (#81145) was identified in the commitments store, where a lack of write serialization causes concurrent updates to silently overwrite each other, leading to lost dismissals or duplicate deliveries.
Key Themes
1. The "Fragile Beta" Experience
There is a clear pattern of "regression-fix-regression" cycles. For example, the claude-cli backend has seen multiple registration failures (#72576), and the thought_signature 400 error for Gemini has reappeared in 2026.4.25 after being supposedly fixed in 2026.4.24 (#72879).
2. UI/UX Friction and Inconsistency
Multiple issues point to a lack of standardization in the CLI and Web UI:
- CLI Conventions:
openclaw sessions listis the only list-style command that rejects thelistsubcommand (#81139). - Web UI Layout: Users on ultrawide displays report a degraded experience where forms stretch across the entire viewport, separating labels from action buttons (#72772).
- Control UI Flashing: The webchat interface briefly flashes the login gate during tab re-focus due to an immediate
connected=falserender (#72500).
3. Memory and Dreaming Pathologies
Issues with the memory-core plugin suggest that Dreaming artifacts are not being written despite successful cron reports (#72546). There are also reports of "compaction loops" where stale transcript usage triggers repeated, unnecessary compactions (#81178).
Action Required
Critical Priority
- Fix Codex Module Resolution: Resolve the
MODULE_NOT_FOUNDerror in the@openclaw/codexpackage to restore the primary OpenAI route. - Address Event Loop Blocking: Move blocking HTTP calls (especially in Telegram startup) and heavy embedding searches (
memory_search) to worker threads or non-blocking queues to prevent gateway hangs. - Implement Secret Redaction for
exec: Add a redaction pass to theexectool output to prevent internal secret exposure. - Patch Auth Bypass: Secure the
shouldSkipLocalBackendSelfPairinglogic to prevent unauthorized pairing bypasses.
High Priority
- Fix Commitments Store Serialization: Implement a writer queue for
commitments.jsonto prevent data loss during concurrent updates. - Resolve Gemini 0-Token Stalls: Investigate the intermittent stalls in
gemini-3.1-pro-previewisolated sessions (#80714). - Repair Telegram Group Delivery: Address the regression where auto-replies are not delivered to group topics (#81207).
Medium Priority
- Standardize CLI Subcommands: Add
listas an alias foropenclaw sessionsto match other parent commands. - Improve Web UI Layouts: Implement max-content widths for ultrawide display support.