HostManager failed to remove container

Odd problem today (finally resolved at 2024-01-18 14:30:06 UTC) when ACU agent was misbehaving, and hostmanager "down" could not kill it (twice, though first attempt was brief). HM logs show:

2024-01-18T14:23:51+0000 start called for update
2024-01-18T14:23:51+0000 update:140 Status is now "starting".
2024-01-18T14:23:51+0000 update:140 Status is now "running".
2024-01-18T14:23:51+0000 update:140 Update requested.
2024-01-18T14:23:51+0000 update:140 Status is now "done".
2024-01-18T14:23:52+0000 manager:0 Requesting termination of ACUAgent:acu
2024-01-18T14:23:57+0000 manager:0 Agent instance ACUAgent:acu refused to die.
2024-01-18T14:23:58+0000 manager:0 Detected unexpected session for ACUAgent:acu (probably docker); it will ...
2024-01-18T14:24:03+0000 manager:0 Requesting termination of ACUAgent:acu
2024-01-18T14:24:08+0000 manager:0 Agent instance ACUAgent:acu refused to die.
2024-01-18T14:24:09+0000 manager:0 Detected unexpected session for ACUAgent:acu (probably docker); it will ...
2024-01-18T14:24:13+0000 manager:0 Requesting termination of ACUAgent:acu
2024-01-18T14:24:19+0000 manager:0 Agent instance ACUAgent:acu refused to die.
...

But running docker-compose rm --stop --force ocs-acu, as ocs user, did bring down the container, without any issues. (That command should be exactly what the HM runs.)

I have no useful analysis yet...

Start with the HostManager shutdown path that runs docker-compose rm --stop --force ocs-acu, then compare it with the logged termination attempts and the successful manual command. Reproduce the ACU agent failure if possible; done means HostManager reliably removes the container when the agent refuses to terminate.