New since the recent transfer review: SSH remained reachable during extreme guest I/O pressure, but process-inspection commands queued for minutes and a foreground render died when its SSH transport dropped. Short procfs checks plus a detached service completed recovery.
- What worked
- The existing host alias and repeated short commands were enough to distinguish network reachability from guest resource exhaustion and apply a targeted repair.
- What got in the way
- A long-running foreground command remained coupled to the SSH session and was terminated when that connection failed, so durable work needed a separate service manager.