Am running 7.6.33 64 bit in a W8.1, W10 environments without issue, and FAIK, you're the first to report the matter, and it going across multiple sciences. Pretty sure the correlation, of it being post-migration related, is not causation. Usually getting stuck units to run again is by stopping the tasks in question for 30 seconds (With 'Leave non-GPU tasks in memory while suspended' off), and then resuming them again. You may verify in task manager these task(s) have left memory, they must, else they remain stuck. Certainly what you did to resolve the matter has erased all traces of a culprit, if it were any piece of BOINC.

When a task finishes, it wishes to do a little housekeeping, zipping up result files and such, the question then arising if something on the host is blocking that. E.g. check the security software, noting WCG swapped IP address during the migration, but you don't seem to have upload/download problems. Strongly recommend to set a scanning exception in the AV for the BOINC data directory and it's subs [Is Sandboxed], usually C:\ProgramData\BOINC.

As for the uninstall / reinstall, it's important to boot between these two steps, as special 'limited rights' boinc accounts are being created by the installer as part of the sandboxing. Uninstalling and then not booting does not remove those from Windows memory i.e. potentially still a polluted environment.

Momentarily can't think of other things to check.

Stranger things happen, but science app processes getting orphaned is the strangest. My recommendation stands on setting scan excludes on the BOINC data dir, which does not stop in-memory scannin. Whilst, suspending the stuck tasks through BOINC Manager to see with LAIM off, would have proven if the client was still in control of those processes, without having to exit BOINC.

Right now the immediate issue is solved, but not the root cause of why they get stuck, but am sure it's something in your local environment. Report back is these stuckers return.