We ended up throwing the SM server on our VMware ESX 3.5 cluster with HA, DRS, and VMotion. All 8 of the 32 GB RAM, dual quad core 2.5 Xeon servers in the cluster have redundant iSCSI connections through 2 switches to our 5 LeftHand NSM2120 clustered SANs. It's a total of 64 processing cores, 160 GHZ, and 256 GB of RAM. This works extremely well. Unfortunately it's a $250k cluster that obviously runs a lot of other stuff and is serious overkill for just failover email.
You could get by with something like the VI3 Standard HA Acceleration kit. It comes with licenses for 4 processors (2 dual processor servers, doesn't count cores), Virtual Center to manage it with up to 3 servers, and HA. It won't move failed servers on the fly or automatically move virtual servers to hosts with the most resources, but if you have a failure it will reboot the virtual machine on the other box. For hardware, you would need a minimum of 2 similar servers, and an iSCSI server that will work with VMware. You're looking at about $7500 for the licensing and mandatory support on the VMware. It could be a good solution if you have multiple web or email servers that you need failover for that is relatively easy to implement.
I don't know if this would be in your price budget, but I have tested and run it in both of the configurations I have mentioned and know it works.