An unknown failure in resolving the keytab cause test failure twice (BF-16430), but I have been unable to reproduce the failure after running the test many thousands of times. The best course of action might just be to add some additional tracing around the keytab check, such as printing out the contents of the keytab and credentials cache before performing the assert. This way, we can have some more information if it happens again.