I don't know if it's the latest version of Parallels Desktop 16.10 (48950), or the fact that I changed the Windows VM configuration from using the Parallels hypervisor to using the Apple hypervisor, but I've been running the virtual machine for five full working days now without the video.metal=0 boot flag, and it has not generated the kernel panic crash at all during that time. I may be lucky, but I would previously have had two or three crashes at least. So perhaps the issue is fixed, or it isn't triggered when using the Apple hypervisor?