Windows Server 2016: How to protect the stability of host fabric in Hyper-V with high load

We have a server running Windows Server 2016 with 4 physical cores (8 hyperlinks) and Hyper-V installed.

Now, let's say that there are two virtual machines, each configured in Hyper-V to have a certain number of X cores. Let's say that each of those virtual machines is under a very high load: 80% to 100% CPU load.

If I want to safely preserve a margin for the host to remain healthy so that I can use the remote desktop on the server to manage it, what is the maximum number of cores that I can assign to each VM in Hyper-V? Should I leave a "free" kernel so that the host remains stable / accessible? If so, can I leave a free partial kernel instead of a full kernel?

Is there a separate configuration to specify the amount of CPU that must be kept for the host machine?

Also, in Hyper-V, does a core mean a physical core (of which there are 4 in this example) or a logical core (of which there are 8 in this example)?

Any help with this is very much appreciated.