You shouldn't have to clear ARP tables etc., when a VM moves as ESX takes care of things by sending a RARP (Reverse ARP) to update the switch tables.
In the explanation of the problem you mention ""for HA and LNB". Should this actually be NLB i.e., Microsoft Network Load Balancing? If so you might want to take a read of the VMware Technical Note Implementing Microsoft Network Load Balancing in a Virtualized Environment.
Regards