I've had to do this a couple of times and always ended up with a working system in the end.
As a fall back option (although I've never had to use it) I have a backup engine VM running completely outside of oVIrt (ESXi host in my case). Then if the hosted_engine deploy fails for any reason you can restore onto the backup vm as a temp solution while you work through the hosted engine deploy issues.
A few things that come to mind: -
* You will need a dedicated gluster volume for hosted_storage and it needs to be replica+arbiter.
* Make sure you put the cluster in global maint mode before performing the engine backup, I recall having issues with the restore when I didn't do that.
* Migrate all other VMs off the host running Engine before doing the backup. This will be the host you will restore onto.