With EVA, check if at the moments you are having the problem you are not havind queue full errors on the vmkernel.log, like the ones shown on the KB http://kb.vmware.com/kb/1008113
H:0x0 D:0x28 P:0x0 Valid sense data: 0x## 0x## 0x##
This is an indicative that the EVA controller queue is being full on the heavy load moments. You will need to call HP to know how to tune this on the ESX servers (and additional server you may have connecting to it).
I have seen some problems related to this on EVA environments, so checking won't hurt 