Thanks, Quinn. I read thread 848542 and understand your recommendation: use normal shutdown/stop paths and rely on Virtualization to clean up after an unexpected client exit, without an additional crash-cleanup mechanism. My remaining question concerns safe reuse of the VM’s backing disk. Is there a supported point observable by a surviving process at which guest execution and new I/O have ceased, and previously submitted I/O can no longer access that disk? Does waitpid completion for the owning process establish this ordering, or is there a separate public completion signal? I am asking about resource reuse under normal OS operation, not guest filesystem durability or absolute deadlines during kernel/hardware failures. If this completion boundary is unspecified, that clarification would help.