The beauty of the UCSM is that it is so flexible like that, in that the same physical blades, can provide configurations that exceed the max number of HBA and NICs that any of the OSs support, or have just a couple of vHBAs/iSCSI BFS NICS, NICs in a simple config.
At a minimum you want to have ether 2 vNICs (1 side A, 1side B) in an active/active OS team, or you can create a single VNIC with fabric failover.
Different OSs can have different best practice requirements, as can the individual use (iSCSI, NFS, Vmotion, tend to like single vNICs, and let MPIO handle storage load balancing, redundancy), vs guestVM traffic that can benefit from teamed nics.
If you have existing Esxi hosts, redesigning or consolidating the number of vNICS/VMNICs presented to the Esxi host, can cause major VMnic re-numbering/ordering, that often requires reinstalling the ESXi OS to straighten out.
Thanks,
Kirk...