This is for a customer.
Goal of this spike is to undesrstand how foreground deletion can be triggered - a client and server side appraoch must be validated:

Something that I have certainty on that it will work:

$ kubectl delete  --cascade='foreground' … 
# Returns once VM and VMI and launcher pod are gone, and not before

But this foreground deletion depends on the client/caller.

The other approach:

A possible alternative is to create a VM object as usual, but with the following finalizer set:

kind: VM (not VMI)
…
metadata:  # VM level
  finalizers:
    - foregroundDeletion
spec:
  template:  # VMI level
    metadata:
      finalizers:
        - foregroundDeletion
…

This VM object can then be deleted with

$ kubectl delete (no --cascade flag!)
# Returns once VM and VMI and launcher pod are gone, and not before

and should still behave like "kubectl delete --cascade=foreground"
This method is server / callee sided, and thus client independent, and what the custoemr is asking for.

Scope of this sprint is to validate both approachs for foreground deletion and open a PR against the kubevirt user-guide to clarify the situation.