The specific assertion failure seems to indicate that the test did not start with a clean environment.

The revision number it returned was 5 but your test case expected 1. I'll confirm locally whether there is a bug, but that would ultimately be my first instinct to check that a prior test didn't leave some obscure objects in the database that tainted the execution of this test for any reason. I've seen this be a problem in the Envers test suite in the past on dialects where we specifically need to clean the database but a fatal error prevented.