One of the biggest, and hopefully easiest, tests that we should add is verification of the priority of the node holding the VIP. In the bug that prompted this work we would have noticed the problem if we had been looking at priorities. In the bad clusters the priority only ever reached 45, not the normal 65 it was supposed to.

It's possible we can cover this as part of the timeline tests, but I wanted to track this separately just in case we need an explicit test for this.