I initiated a workflow and monitored it for the first 10 minutes, where it appeared to function normally. I then queued a subsequent task and left the agent running. Upon returning 6 hours later, I discovered that:
-
The agent completely froze shortly after the initial 10 minutes, generating zero progress or output for nearly 6 hours.
-
The queued task was never triggered, despite the pipeline being locked and active the entire time.
-
Severe Quota Waste: This freeze needlessly burned 20% of my 5-hour limit on absolutely zero output or progress.
-
This resulted in a massive waste of execution time, quota/credits, and completely disrupted my project schedule.
To get it to resume working, I have to manually interrupt it and tell it to continue. After doing so, it enters a loop with the following error:
Stop hook blocked termination: You are still working toward the user's goal. Do not stop until the task is fully complete. Time spent so far: 9h 12m Continuation: #1
Then, I have to interrupt it a second time and tell it to continue again before it actually proceeds

