> When you log into all of your nodes after a network snafu and discover that every single node has attempted to perform a full state transfer, involving the destruction of the data directory...

I support customers that are using Galera at Red Hat and while we have seen lots of network snafus and situations bringing individual nodes back online, I've never seen all nodes attempt to state transfer each other like that nor have I ever seen a data directory "destroyed". The situations with Galera in 100% of cases involve operators acting too hastily and not understanding what's going on as they do it.