Experience shows that when Native Persistence is enabled, it is better to explicitly set ConsistentIDs than use the autogenerated ones.
First, it simplifies managing the baseline topology. It is much easier to manage it via control.sh when the nodes have stable and meaningful names.
Second, it helps to avoid certain shoot-yourself-in-the-foot issues. E.g. if one loses all the data of a baseline node, when that node is restarted it doesn't have its old autogenerated consistent ID - so it is not a part of the baseline anymore. This may be unexpected and confusing.
Finally, having explicit consistent IDs improves the general stability of the setup - one knows what the the set of nodes, where they run and what they're called.
All in all, it seems beneficial to urge users to explicitly configure consistent IDs. We can do this by introducing a warning that is printed every time a new consistent ID is automatically generated. It should also be printed when a node doesn't have an explicit consistent ID and picks up one from an existing peristence folder.
This message was sent by Atlassian JIRA