Observed this behavior in Qwen 3.8 27B (full-precision) serving via vLLM ver 0.28.0
Logs:
....
time=2026-09-03T06:54:36.327Z level=DEBUG msg="OpenAI chat completion stream created successfully" model=Qwen/Qwen3.8-27B
time=2026-09-03T06:54:36.327Z level=DEBUG msg="Processing stream" agent=orchestrator model=verda/Qwen/Qwen3.8-27B
time=2026-09-03T06:54:36.334Z level=DEBUG msg="[STORE] AddMessage" session_id=d64c026e-4654-4137-9e33-b3c02621e5dc message_id=42 role=assistant agent=orchestrator
time=2026-09-03T06:54:36.334Z level=DEBUG msg="[PERSIST] Created streaming message" session_id=d64c026e-4654-4137-9e33-b3c02621e5dc message_id=42 agent=orchestrator
time=2026-09-03T06:54:36.453Z level=DEBUG msg="Stream processed" agent=orchestrator tool_calls=0 content_length=1137 stopped=true
time=2026-09-03T06:54:36.453Z level=DEBUG msg="Added assistant message to session" agent=orchestrator total_messages=42
time=2026-09-03T06:54:36.454Z level=DEBUG msg="Processing tool calls" agent=orchestrator call_count=0
time=2026-09-03T06:54:36.454Z level=DEBUG msg="Conversation stopped" agent=orchestrator
---- below 2 logs are after I pressed "esc" and interrupted the conversation (and that worked fine).
time=2026-09-03T06:54:36.454Z level=DEBUG msg="Executing hooks" event=session_end session_id=d64c026e-4654-4137-9e33-b3c02621e5dc count=1
time=2026-09-03T06:54:36.454Z level=DEBUG msg="Executing on-user-input hooks" context="stream stopped"

Observed this behavior in Qwen 3.8 27B (full-precision) serving via vLLM ver 0.28.0
Logs: