openclaw - 💡(How to fix) Fix CLI commands ~14s cold-start regression after 2026.5.12 update [1 pull requests]

Summary

All CLI commands are taking ~14 seconds to complete after updating from 2026.5.7 → 2026.5.12 on Linux. The gateway itself is healthy — this is pure CLI-side startup overhead.

Steps to reproduce

time openclaw health

Output

Telegram: configured
Gateway event loop: ok max=35ms p99=25ms util=0.023 cpu=0.022
Agents: main (default)
...

real    0m14.555s
user    0m10.658s
sys     0m1.206s

~10.6s of user CPU before the command completes. Gateway event loop is fine.

Environment

• OpenClaw: 2026.5.12 (was 2026.5.7)
• Platform: Linux 6.17.0-20-generic x86_64
• Node: v22.22.2
• Install method: pnpm (npm-global)
• Gateway mode: local systemd user service, port 18789

Additional context

• The codex plugin (@openclaw/codex) was installed for the first time during this update.
• Immediately after the update the gateway also had a severe event-loop spike at startup (eventLoopDelayMaxMs=117037ms, util=0.989) that lasted a few minutes. That has since resolved — the gateway is stable — but CLI cold-start remains at ~14s.
• openclaw completion --write-state takes >120s (was timing out at the update script's 30s cutoff).
• Pre-update (2026.5.7), CLI commands were fast. No baseline timing captured unfortunately.

The 10.6s of user CPU on the CLI process suggests heavy module loading at startup, possibly from the new Codex plugin being pulled into the CLI's require graph.