Sounds like you're on the right path.
Yes, if you lose one host, HA will bring those VMs online on the other host.
Bringing up anything replicated by vSphere Replication will take a manual process.
What's I'm hearing here though is "we've got a very low budget NFS server, and we're engineering ourselves some form of availability". I've walked into plenty of environments like this, and I have to blunt, you are almost certain to engineer yourself downtime due to your own complexities and unusual implementation, whereas a quality SAN would almost certainly never go down.