LHC@home

Often the VM spontaneously turns off after 20-100 minutes working, then the BOINC manager shows the status "VM job unmanageable". Also, the VM can suddenly disappear, then "Computation Error" occurs.
VirtualBox 5.2.0 r118431, Windows 10 Home x64 1709

In your logs , you can see the mention :

BOINC will be notified that it needs to clean up the environment.
This is a temporary problem and so this job will be rescheduled for another time.

You have to open virtualbox manager and delete the virtual machines which appears with a red mark ( that means : unreachable, the vboxwrapper can't link anymore with virtualbox)

2017-10-23 23:25:53 (4840): ERROR: Vboxwrapper lost communication with VirtualBox, rescheduling task for a later time.

It may happen when there is an error produced because the vm has not the time to save or stop its state properly.
So , to avoid serial errors ,

2017-10-23 23:31:53 (3004): Error creating VirtualBox instance! rc = 0x80004002

you have to sometimes manually delete the vms in virtualbox manager by right-clicking on it (but only the bad ones in a bad state).

If you have trouble with atlas tasks , because ,you seem to abort them volontarily , you can go to your web preferences and uncheck atlas jobs in your applications.