As of 1.7, migrations have been split into three different migration types:

Intercluster full workload stage and migrate
In-place storageclass conversions
State transfer

With the way that state transfer functions currently, we need to ensure that
the repeated stage + final cutover fits the original use-case cleanly of
integrating state transfser with applications that are managed outside the cluster
by an external CD system.

Imagine you have a workload on the source cluster that is versioned and ready
to be deployed to a target cluster using the gitops flow, but you need a solution
to help you get your state to the target cluster. With 1.7, I should be able
to do the following:

* Run a stage state transfer that gets all of my PVCs to the target namespace
without anything else. This gets the bulk of my data to the target without having
to accept downtime for my source app.

* Run a cutover state transfer that is effectively exactly a stage transfer (i.e.
incremental data transfer into the pvcs that are alrady present on the target
from the inital stage), but it *quiesces the source workloads first*

* I use my CD system to deploy into the namespace with my data waiting for me
in my PVCs

Right now cutover migrations will actually bring over application resources, when
they should only be doing data transfers.