Is this the same cluster that is experiencing the issue described in the
thread "intermitent failures led to node restart"? If so, I would be
surprised if this is not the same issue being experienced at different
times (one before a node joins the quorum and one after).