Whenever I shut down Windows 10, and Virtualbox is in use (eg an LHC task), Windows says "Virtualbox has active connections". I can force it to stop, but if it's a remote computer this is annoying or very difficult to do, and it doesn't seem right to force stop a running project. If I wait, Windows eventually gives up and goes back to the desktop. Sometimes if I try to shut down again, it works, but sometimes I get the same error message again. I've done a Google search and this seems to have been a problem for years. Is this a Virtualbox problem, a Windows 10 problem, a Boinc problem, or an LHC problem?

Virtualbox wasn't designed to have BOINC control it in such a fashion. You will also find if you pause BOINC that virtualbox does not pause. Also you will find vitrualbox uses all your cores. Remember it is an O/S.

Windows won't know exactly what VirtualBox is doing, but it's the LHC virtual machine which will have the connections open to CERN. Ask about it at LHC - BOINC advice can't go that deep into an individual project.

Virtualbox wasn't designed to have BOINC control it in such a fashion. You will also find if you pause BOINC that virtualbox does not pause. Also you will find vitrualbox uses all your cores. Remember it is an O/S.

Both BOINC and CERN have put in a lot of effort over 10 years to ensure that information can flow out of VB to BOINC, and control signals can flow inwards.

Virtualbox wasn't designed to have BOINC control it in such a fashion. You will also find if you pause BOINC that virtualbox does not pause.

When I pause Boinc, the CPU stops completely, I just paused a 24 core machine in Boinc, 6 of the cores running LHC Theory tasks under Vbox. The CPU usage went to zero almost immediately.

Also you will find vitrualbox uses all your cores. Remember it is an O/S.

Not sure what you mean by that. It uses what Boinc expects it to. LHC Theory uses 1 core per task/VM. LHC Atlas uses up to 8, depending on my settings, but this is displayed in Boinc, Boinc knows how many cores are allocated to that task.

On my boxes zero isn't zero but 0.xxx. VB never quite goes to zero on its own. Every O/S, and VB is a guest O/S on Doze, always has some stay awake tasks.

As to the active connections issue, while that is IP related I think that is the 127.0.0.1 loopback that is active for inter-process signals and not traffic to outside.

No matter, when BOINC asks VB to shut, which is does when Doze tells BOINC to die from your shutdown, VB is not killing itself. I'm not sure why, but VB is a privileged process and BOINC is not and it may be the signal is blocked by Doze security. Or Doze kills BOINC before it has a chance to ask VB to quit. Or perhaps the reverse and Doze asks VB to quit but it still has active science which it can't kill until it informs BOINC who Doze may have killed by then. Think checkpoints.

PS, none of this BS happens on Linux or Mac, just Doze.

As to VB using as many cores as it wants, it does. While the LHC science app only uses one or however many it declares, every time that science app makes a system call to the O/S, VB in this case, VB uses as many cores as it wants to service that request. VB is not bound by BOINC in this manner. Nor is any O/S and don't forget VB has to call Doze to do the request so Doze will wake up some cores too.

My only suggestion is to pause BOINC, then quit it telling it to quit science. Once that is done, then restart or shutdown Doze. If you get the message then, you know it is VB/Doze issue.

Do you know a way to tell Boinc on a remote computer to quit? I use Boinctasks on this one to control the others, but I don't think I can get remote Boincs to close completely, just pause.

If BoincTasks can see the remote machine, so can BOINC Manager - same hostname and password for all variants of Controlling BOINC remotely.

File menu, "Shut down connected client", certainly works when sent to a remote computer.