Improve Nifi Cluster Behavior In Down Scenarios

Improve Nifi cluster behavior by allowing nodes to receive flow changes when not all cluster nodes are online, so long as the node(s?) receiving the changes are registered/active with Zookeeper. Then when require down nodes to update to latest flow version with cluster before cluster re-join is completed. This would probably require external/non-embedded zookeeper to insure valid quorum. For any flow files already queued up on a node, process them on the old flow prior to changing to the new flow and make sure flow is current before starting any processors consuming data from outside systems.
This was born out of a situation where one of our pre-prod cluster nodes has been down for nearly a week, and the cluster being read-only until all nodes are back online has prevented any further development work from being done.