Hi,

yes you're right, the issue might be different from what I suspected.
I also must pinpoint that I don't like the filter in 5.3.2 at all: it might be cause performance issues on big meta-models with thousands of entity classes.
Moreover I don't understand at all, why hibernate should automatically call addEntity with an completely arbitrary alias with name "alias1".

I made a pull-request with the required testcase(s), I also commented out the problematic code in AbstractSharedSessionContract.java#L934
in order to make the tests pass.