Bug type
Regression (worked before, now fails)
Beta release blocker
No
Summary
Summary
Gateway stalls for ~2–3 minutes at exact half-hour boundaries (:29-:32 and :59-:02).
During stall, Telegram and Control UI both become unresponsive.
Environment
- OpenClaw: 2026.4.2 (d74a122)
- OS: WSL2 Ubuntu 24.04
- Node: 22.22.1
- Model setup: anthropic/claude-opus-4-6 main, local llama fallback
Symptoms
- Control UI messages briefly appear then disappear / no processing- Telegram timeouts or delayed responses
- Gateway health endpoint often still reports OK during stall
What I already tried
- Disabled heartbeat / changed heartbeat cadence
- Verified cron jobs are empty
- Reinstalled OpenClaw
- Disabled custom helper services (sentinel/dashboard/fake-ollama/llama)
- Rebooted host
- Issue still reproduces on this machine
Timing pattern
- Reproducible since April 3
- Clock-precise windows around :29/:59 every hour
Steps to reproduce
nothing just wait for the time
Expected behavior
this happened starting around April 3rd midnight when i was trying to figure out how to get gemma 4 working on my system and i also updated my openclaw same time cause i realized it was on a old version of it.
Actual behavior
not supposed to freeze and stall.
OpenClaw version
Current newest version
Operating system
Windows 11
Install method
No response
Model
Anthropic Opus
Provider / routing chain
Provider / routing chain: Inbound: Telegram bot + Control UI Gateway runtime: OpenClaw on WSL2 Ubuntu 24.04 Agent routing: main agent (default) Model routing: primary Anthropic, fallback local llama-server (OpenAI-compatible localhost endpoint) Outbound: Telegram + Control UI Note: fake-ollama shim service existed previously on localhost:11434 and was disabled during troubleshooting.
Additional provider/model setup details
No response
Logs, screenshots, and evidence
Impact and severity
stalls for 3-5 mins every 30 mins at around :28-:32 and :58-:02 at every hour and sometimes longer depending on what it was doing at the time. Been dealing with this for days and its a nightmare especially for what my agents are made for.
Additional information
No response
Bug type
Regression (worked before, now fails)
Beta release blocker
No
Summary
Summary
Gateway stalls for ~2–3 minutes at exact half-hour boundaries (:29-:32 and :59-:02).
During stall, Telegram and Control UI both become unresponsive.
Environment
Symptoms
What I already tried
Timing pattern
Steps to reproduce
nothing just wait for the time
Expected behavior
this happened starting around April 3rd midnight when i was trying to figure out how to get gemma 4 working on my system and i also updated my openclaw same time cause i realized it was on a old version of it.
Actual behavior
not supposed to freeze and stall.
OpenClaw version
Current newest version
Operating system
Windows 11
Install method
No response
Model
Anthropic Opus
Provider / routing chain
Provider / routing chain: Inbound: Telegram bot + Control UI Gateway runtime: OpenClaw on WSL2 Ubuntu 24.04 Agent routing: main agent (default) Model routing: primary Anthropic, fallback local llama-server (OpenAI-compatible localhost endpoint) Outbound: Telegram + Control UI Note: fake-ollama shim service existed previously on localhost:11434 and was disabled during troubleshooting.
Additional provider/model setup details
No response
Logs, screenshots, and evidence
Impact and severity
stalls for 3-5 mins every 30 mins at around :28-:32 and :58-:02 at every hour and sometimes longer depending on what it was doing at the time. Been dealing with this for days and its a nightmare especially for what my agents are made for.
Additional information
No response