A customer support ticket I got today started out with perceived latency, ended up with a rescan of HBAs that dropped a datastore, & left me at my wit’s end. After banging my head against the wall all day I finally broke down & called VMware Support. According to the tech, using Microsoft Clustering can greatly increase the likelihood of SCSI reservation conflicts in an ESX cluster, which can result in the symptoms I was seeing. This is the only cluster we have where MSCS is even running.
Luckily for me, this is a well-known problem with a well-documented solution:
(this cluster consists of ESXi 4.0 U1 hosts)