We had an instance where 4 out of six blades lost connectivity to 1 LUN. For some odd reason, this caused our ESXi hosts to become non-responsive although the VMs stayed up. The solution was to remap the LUN from our fibre storage array (Compellent) ...