Had some problems in a customer environment. we had Cluster volumes that went offline from time to time.
there where a three node hyper-v cluster and a HP P2000 SAN, with eight NIC:s and a Veeam backup solution.
All hosts, including the Veeam backup server, had fully redundant connections to the SAN. this meant that all servers had 2 NIC and MPIO x 8 NIC on SAN connections. The SAN had 7 LUNs which meant that in Hyper-V cluster that meant 4 hosts x 2 paths x 8 SAN-NICS x 7 LUNs = 4*2*8*7=448 persistent connections. The HP P2000 has a reservation table for max 128 connections.
As a conclusion: to have a preservation table ¼ the connections from a cluster makes a bad day. We changed it to use 4 hosts x 2 paths x 2 SAN-NICs x 7 LUNs which is 112 and lower than 128.