Pre 3.5.0 zookeeper reconfiguration of a running cluster was also much harder - that was a significant discussion point on Kubernetes when we had the etcd vs (a

Pre 3.5.0 zookeeper reconfiguration of a running cluster was also much harder - that was a significant discussion point on Kubernetes when we had the etcd vs (anything) discussions early after open sourcing. I still think etcd total ordering over history also made reasoning about changes in the system while we were writing the first versions of the controllers and caches and list-watch loops. ZK had partial order, and I was leery of that at the time.