This was reported and discussed in another thread although I forget the title. The issue is with the boinc manager not being able to handle the sheer numbers of work units being generated by HCMD2.

Something has gone wonky and you'll need to just abort that work unit. Expect the problem to keep reappearing.

It is also far more common on Linux systems. There is no known fix and this will need to be addressed by the boinc programmers as WCG can do nothing to correct that software.

As a short term fix, we have deployed a change that limits a computer to having at most 60 hcmd2 workunits on the client at a time. This should stop the issues related to the client failing to properly managed large numbers of workunits at once. We will dig into why they were sized so small.