Some application utilize the same table definitions across a several schemas or separate database instances to isolate customer data.
A pure JPA solution to this problem would be to create a separate EntityManagerFactory for each schema & database instance. This approach scales poorly, and results in a lot of duplicate processing time, and memory used.
As an alternative we could provide the ability to specify the connectionFactory to use when the EntityManager is created. Effectively this allows the EntityManagers to share a MetaDataRepository but communicate with different databases.
There are some limitations to this support, obviously the table definitions must be identical for each database used, if the OR mappings include a schema that schema must exist (and be writeable) in each database come to mind, but in the long run for applications which follow this specific pattern this function could be a large benefit.