Thank you for your reply, I thougth of this myself but the fact that the two simple cases worked threw me off... The simple cases only work coincidentally, because the registers get fully overwritten

Hi, using Java, I created a list containing the registers. I had no issues running the simple tests neither in an IDE nor in the Web. However, when trying to submit, only the first executed test case works, all others had issues, because the register list of the last test execution seemed to be kept in memory somehow. Adding a .clear() solved this, but IMO, this shouldn´t be necessary.
How can this possibly happen?

Sorry, I agree that everything is as specified. Somehow I must have mixed up some information with my own thoughts...

The test cases contain randomly generated String patterns.

Studying the given test cases for the Python version of this kata, it seems as if none of the tests actually check for uppercase vowels as described in the introduction.