Bug type
Crash (process/app exits or hangs)
Beta release blocker
No
Summary
After upgrading to 2026.4.15, the gateway can freeze silently mid-session with no error — webchat and WhatsApp stop responding — and only recovers when WhatsApp emits a 408 and the health monitor restarts the channel.
Steps to reproduce
NOT_ENOUGH_INFO
Expected behavior
Gateway remains responsive between WhatsApp disconnects, as it did before 2026.4.15.
Actual behavior
Gateway stops processing requests mid-session. No error in logs, no crash. Last ws ⇄ res entry at 23:19:40, followed by 9 minutes of silence. Recovery only triggered by WhatsApp 408 at 23:28:30.
OpenClaw version
2026.4.15 (041266a)
Operating system
Windows 10 (10.0.19045 x64)
Install method
npm global
Model
openai-codex/gpt-5.4
Provider / routing chain
openclaw → ChatGPT OAuth (OpenAI Codex)
Additional provider/model setup details
No response
Logs, screenshots, and evidence
23:19:40 [ws] ⇄ res ✓ node.list 5150ms ← last normal response
← 9 minutes of silence
23:28:30 [whatsapp] Web connection closed (status 408). Retry 1/12 in 2.01s…
23:28:30 [health-monitor] [whatsapp:default] health-monitor: restarting (reason: disconnected)
23:28:42 [whatsapp] Listening for personal WhatsApp inbound messages.
23:29:21 [whatsapp] Inbound message ← fully recovered
Impact and severity
Affected: WhatsApp channel on Windows
Severity: High (blocks all responses until recovery)
Frequency: Intermittent
Consequence: Agent unresponsive on both webchat and WhatsApp until 408 triggers health monitor restart
Additional information
Before 2026.4.15, same scenario caused a hard crash (Unhandled promise rejection: Error: Agent listener invoked outside active run). The update fixed the crash but left this silent freeze. openclaw health --verbose was not captured during the freeze — will add if it reproduces.
Bug type
Crash (process/app exits or hangs)
Beta release blocker
No
Summary
After upgrading to 2026.4.15, the gateway can freeze silently mid-session with no error — webchat and WhatsApp stop responding — and only recovers when WhatsApp emits a 408 and the health monitor restarts the channel.
Steps to reproduce
NOT_ENOUGH_INFO
Expected behavior
Gateway remains responsive between WhatsApp disconnects, as it did before 2026.4.15.
Actual behavior
Gateway stops processing requests mid-session. No error in logs, no crash. Last ws ⇄ res entry at 23:19:40, followed by 9 minutes of silence. Recovery only triggered by WhatsApp 408 at 23:28:30.
OpenClaw version
2026.4.15 (041266a)
Operating system
Windows 10 (10.0.19045 x64)
Install method
npm global
Model
openai-codex/gpt-5.4
Provider / routing chain
openclaw → ChatGPT OAuth (OpenAI Codex)
Additional provider/model setup details
No response
Logs, screenshots, and evidence
Impact and severity
Affected: WhatsApp channel on Windows
Severity: High (blocks all responses until recovery)
Frequency: Intermittent
Consequence: Agent unresponsive on both webchat and WhatsApp until 408 triggers health monitor restart
Additional information
Before 2026.4.15, same scenario caused a hard crash (
Unhandled promise rejection: Error: Agent listener invoked outside active run). The update fixed the crash but left this silent freeze.openclaw health --verbosewas not captured during the freeze — will add if it reproduces.