That has to be annoying to fix. It's one thing when it's "just" your own cluster and something explodes in your face, it's a quite another when it's other people persistent volumes and you can't just assume that they'll be able to redeploy, because you have no idea what is in those storage claims.
Glad it's not me having to deal with this, Kubernetes is still a black box of magic in many respects and there's probably more than a few abstraction down to the actual disks.
Most of the abstraction between disk and container is external to K8s. You can use a single box CPU+disk with K8s if you want, that's just not a configurable product offering, so that's not how any of the cloud providers expose disk. K8s itself is entirely unconcerned with most of the implementation details, it's more a framework for those kinds of things
Glad it's not me having to deal with this, Kubernetes is still a black box of magic in many respects and there's probably more than a few abstraction down to the actual disks.