We have this technique built into our management processes. So when funnies are suspected with the vCLS they can be refreshed at the click of a button. It seems to cure all problems we have encountered where the appliances 'seem' to be sort of functioning but not behaving 100% as expected.
I think VMware are still learning about vCLS and how to make it work better. So it has a feel of being "work in progress".
We have problems with the way vCLS are placed at system startup, because there is not in my opinion enough control mechanisms available to users to control placement. It is possible for example to end up with all your vCLS in one blade chassis, in one rack or in one room. So a major outage of some sort can obliterate all the vCLS for a cluster in one go. This leaves HA recovery at the mercy of the old HA placement engine which still exists inside the hypervisors. This engine is now a little bit broken in that it no longer obeys the advanced affinity rule controls, so if you lose all vCLS there is a chance that HA recovery will be without any affinity rules being applied at all.