I was recently working on spinning up a Keycloak cluster. We're deployed to AWS, can't leverage UDP and wanted to have a discovery mechanism that was easy to use, so tried out JDBC_PING. While I got the cluster running, when shutting down a node I would receive an exception on the console when attempting to remove the node from the cluster. It attempts to do a delete on the row in the database after the DataSource has been shut down.
In my setup, I'm pointing JGroups at the same DataSource as Keycloak, I'm not using the properties approach. I can switch to the properties approach, but I would prefer a single pool rather than multiple connections.