Another implication here... this effectively means a split in EntityPersister#hasCache to account for put and get separately:

1. should put when the entity being persisted is cached
2. should get when the entity or any of its subclasses is cached.