Machine: Early 2013 MacBook Pro running Catalina 10.15.5 (19F101) and Parallels 15.1.4 (47270). Since upgrading to 15.1, I've had both a Windows 10 and a Fedora VM completely lock up, forcing a vm restart or stop, completely trashing whatever was on virtual disk. When the lockup happens, there's a low-level but constant disk read occurring at around 13-15 Mbps from prl_disp_service. A normal file transfer from an external USB 2.0 disk to my internal SDD is well over 35Mbps, so an internal SDD transfer of a large (say 50-70GB virtual hard disk) should be a lot faster process than that. What's weird is that although the disk transfers are low rate, and CPU consumption is also low, the memory usage is off the charts and my computer gets a cursor spin on every click, making it completely unusable: All Finder and disk operations get tied up, which gets further compounded if Time Machine kicks off. Because the maintenance is so slow, drawn out, and resource heavy it, it could take hours to complete, so Time Machine cycle is almost always likely to kick off right in the middle of it. Because there's no warning that this is about to occur, there's no way to tell Time Machine to stop, so the problem gets even worse. My guess is this has to do with periodic vdisk maintenance to either expand or slim it down, which I seem to remember used to be something you could switch off though I can't seem to find where that is anymore. Bottom line: If this is in fact what is happening right in the middle of when a VM is running and there's no alert to the user whatsoever, this is bad. Real, real bad. Even regular backups of the VMs will eventually cause a recurrence of the issue, although it will reduce the negative impact of having to reinstall/reupgrade fewer things later. Nevertheless, it's still an unwarranted pain - please investigate.
Follow-up: I've turned off TRIM on all of my large VMs and will resort to pseudo-manually reclaiming any post-update free space (after Time Machine finishes) to see if this is a viable workaround.
