#latency-issue
Sorted by views, then solution_desc, solution, and root_cause length (desc).
2551 issues
On the same Ubuntu VPS and openai-codex OAuth setup, OpenClaw 2026.4.9 fails to return runtime chat replies in Control UI, WhatsApp, and WeCom, while 2026.3.28 works.
After upgrated to 2026.4.1, the gateway display the following message after input some message in openclaw webpage: 17:48:38+08:00 [agent/embedded] embedded run failover decision: runId=fa92620f-4ad8-40d4-9855-9f82a5d42ac8 stage=assistant decision=surface_error reason=timeout provider=custom-127-0-0-1-8088/NVIDIA-Nemotron-3-Super-120B-A12B profile=- ... and then the LLM running is idle, the openclaw stopped working.
After upgrading to OpenClaw 2026.3.13 on Windows, the ACP runtime backend is unavailable. This appears to be a recurrence of #43997, but with two compounding issues specific to Windows.
Allow vault/exec secret references (e.g. OpenBao, HashiCorp Vault) to be used as the `apiKey` source in `auth.profiles`, the same way they are already supported in `tools`, `channels`, `gateway`, `skills`, and `messages`.
`task-manager` workers can resolve `SKILL.md` from the bundled OpenClaw install path instead of the current workspace path, which causes startup to fail with `ENOENT`.
The `openclaw gateway restart` command sends SIGTERM to the gateway process but does not verify that the gateway successfully restarts. This can leave the gateway down for extended periods with no error indication or automatic recovery.
Running openclaw gateway probe returns "Connect: failed - timeout" even though the gateway is fully functional (HTTP works, WebSocket connections work). Environment: OpenClaw: 2026.3.13 OS: Linux 6.12.62+rpt-rpi-2712 (arm64) - Raspberry Pi 5 Gateway: loopback (127.0.0.1:18789) Steps to reproduce:
The built-in auto-update consistently kills the gateway on macOS with no recovery, leaving the service unloaded indefinitely. The root cause is that `resolveCommandStdio` forces piped stdout/stderr for the update subprocess, creating a fatal dependency on the gateway parent process — which is then killed by `launchctl bootout` during the restart step. This is the underlying mechanism behind several related reports (#54861, #57379, #58041, #40811, #48992) that describe symptoms (service unloaded, multi-hour outages, KeepAlive not respected) without pinpointing why the restart handoff never completes.
I hit two problems on a Linux git install of OpenClaw: 1. `openclaw update` failed with `preflight-no-good-commit`. 2. Host exec works, but auto-exec / approval behavior appears inconsistent: - harmless commands were blocked with `Blocked by allowlist. Need approval to proceed.` - no usable approval popup/UI appeared - `/approve ...` failed with `unknown or expired approval id` - manual allowlisting of binaries made the same commands succeed This suggests: - the updater is hiding the actionable dependency-install reason behind a generic preflight error - the exec engine itself works, but the approval / allowlist / auto-exec flow is broken or inconsistent
OpenClaw 2026.3.11 on Ubuntu VPS. gog CLI works and Google OAuth is successful. openclaw skills check shows gog as ready. But when running in the browser: /skill gog gmail labels list the UI returns: No session found: gog Gateway logs show: sessions.resolve errorMessage=No session found: gog The gog CLI works directly from the VPS.