Failover

They are just two different ways of achieving the same purpose. F5 lists unicast mesh as an optional method, and it can be used when the multicast service is not available for your network. I think while multicast could get you a fractionally faster failover in some situations, a unicast full mesh would be the most robust method when you have multiple blades installed and/or the connecting network device is beyond your control at times of network trouble.

On the other hand, F5 does recommend dedicated trunk/vlan/self IP address for HA failover; the traffic generated by the mgmt operation could have adverse impact to an HA operation.

You can see the HA traffic by this command:

 tcpdump -nni 0.0:nnn -s0 port 1026

.