1) I haven't, but I'd try. I can't think of any reason why this wouldn't work well.
2) Should be the same as Windows bare metal server.
3) SCSI-3 reservations are supported so it should just work as per Windows setup guidelines, I think.
Some things I'd watch out for:
- If your setup is HDD, obviously it won't be "fast" for small IO, so you can limit IO queue depth on Windows iSCSI initiator or NIC. For Hyper-V, as I recall you can also set limits on VMs (e.g. if you have 20 NL-SAS disks, that'd be around 1500 IOPS, so set a low queue depth (4, for example) and on VMs set an IOPS limit per VM (e.g. with 30 VMs, set 100 IOPS limit on each, if you assume 50% would run at max while others are idle). Watch perf counters: https://learn.microsoft.com/en-us/windows-server/administration/performance-tuning/role/hyper-v-server/detecting-virtualized-environment-bottlenecks#s...
- You can use jumbo frames for iSCSI, but you need room for Hyper-V overheads, so use iSCSI frame size slightly smaller than E-Series NIC has (e.g. if E is 9000, try 8800 on iSCSI client NIC)
- As I recall you can't use multiple VLANs on E iSCSI, so use default VLAN