Currently, our cloud tooling configures clustering in the cloud using a replicated cache mode which is suboptimal as it does not scale.

Such cache mode does not scale and is not suited for applications in the cloud that are supposed to be scalable. In this mode Infinispan replicates all entries in the cache to all nodes in the cluster. It doesn't scale, as adding new nodes does not enlarge the cluster capacity.

Moreover, we should relax the isolation level to READ_COMMITTED (from REPEATABLE_READ) and transaction mode to NONE (from BATCH).

Furthermore, this pertains to both EJB and HTTP web session configurations. While at it, we should update the caches to appropriate names: i.e. sessions and beans.

The wildfly-jar-maven-plugin also has a similar scripts to the wildfly-cekit-modules that also need adjusting. The script doesn't appear to set expiration=0 and doesn't seem to reconfigure EJB.