We have deployed Hyper-V last year on CIFS and we have been running into issue with that. One ONTAP upgrade led to lifmgr going into non-responsive state and another was with the domain authentications, rebooting all the VMs. This has led us to think about moving to iSCSI instead of using CIFS. In your opinion what would be the best protocol for Hyper-V and what could be the downsides related to that.
I think that the answer is depend on your circumstances.
If you can get the same (or acceptable) performance and availability with other protocol that been proven or theoretically more fit - that's a win.
Beside MPIO that was mentioned before. I do want to add that by doing that you're moving what now been your point of failure - to another; Windows clustering... in the last few years iv'e seen the CMS process crashes, indefinite WMI leaks (some recovered gracefully, required a VM shutdown or just crashed) and lot's of finger pointing between software and hardware vendor when these go south. yes - the majority is for a single host and not affected the whole cluster, but I did see a few that did.