It would be NICE if VMWare would actually attempt to fix this problem. I have this same problem with about half my hosts at work, where we have in excess of 120 across 6 states. Each time, we have to remove the hosts from the cluster, make a new cluster, an moving the hosts to the new cluster, VMWare's official "fix" for this issue. Well, that's a whole heck of a lot of work, and doesn't work permanently. Every single time I have to reboot hosts, I hope to God that it doesn't happen again, and guess what, it happened again tonight, with a cluster that's had this problem twice before. Before you ask, yes, this cluster is on the most recent patching level of 6.5, so it's current.
It's not fixed. It's been ongoing for over 6 YEARS now, and across 5.1, 5.5, 6.0, and 6.5! (I haven't used 5.0 or 6.7 enough to know if it affects those, but likely.)
Hey, VMWare, FIX THIS.