#🐛 Performance Regression: core-plugin-tools ~16s per run (2026.4.29)

1 messages · Page 1 of 1 (latest)

austere solar
#

Hi all - I first noticed this issue when upgrading from 2026.4.23 to 4.25 (I think).

Really long delays until the first token back from any chats whether through Discord or the WebUI. I downgraded back to 4.23 and thought I'd give 4.29 a try and see if the issue was still there and it is.

I spent significant time trying to troubleshoot without much success. Here's the detail below.

Environment
Ubuntu Linux 6.17.0-19-generic, Node 22.22.2
OpenClaw 2026.4.29 (was faster on 2026.4.23)
Plugins: acpx, browser, device-pair, discord, file-transfer, memory-core, phone-control, talk-voice

Symptom
Every run has ~40-47s startup overhead before the model starts, consistent across all models, sessions, and channels (web UI + Discord):
core-plugin-tools: ~16,000ms ← every single run
system-prompt: ~10,000ms
stream-setup: ~10,000ms
model-resolution: ~4,000ms
auth: ~6,000ms
What was tried
Multiple models (sonnet, haiku, minimax, gpt-5.4) — no difference
Removed dead ollama plugin — no improvement
Cleaned up model catalog — no improvement
Consistent all day; not a cold-start issue

Also observed
Liveness warnings: eventLoopDelayMaxMs up to 24,000ms, eventLoopUtilization=1.0
Multiple [fetch-timeout] events throughout the day

Not a config issue — this appears to be a regression introduced in 2026.4.29. Happy to share full trace logs if helpful.

main craneBOT
#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

foggy river