I always stick with the theory "if its not broken, dont fix it".

If you have a live site working well enough i'd leave it alone until i could be sure the upgrade would be worthwhile. A couple of new functions might sound great but, if they are at the expense of time and aggravation and the disfunction of other parts, it really isn't worth it.
I've got dolphin installed in a test-dev area, so whilst the upgrades can be a rigmarole it doesnt affect me too much.

For the person who paid see more for their site, i fully understand. I've seen a load of clients at work recently where we've been fixing botch website jobs by other people/companies.