This case was about an EBS-backed instance, which is shared storage. Either way, you could avoid centrally shared storage by migrating the backing store first, then the running instance.

In any case, I'm not really trying to criticise here. Just pointing out an engineering trade-off.