I note that this doesn't test with non-ASCII inputs either.

Approved by someone

these warnings were cleared at some point

I've closed this fork.

Fixed in this fork

Added some time ago

No longer an issue.

If there are test cases which always expect some specific answer consistently in the same invocation, it makes test cases suspectible to counting. Users can use a global variable to count invocations of their solution, and return true, false, or None or some other predictable result for every expected n-th invocation.

This problem might not be especially relevant for some kata, the most affected kata are the ones which expect bool as an answer, but the general practice is that for random tests, the expected results (and their order) should be unpredictable This can be achieved for example by generating a list of inputs for all tests, and then shuffling the list.

How does having separate tests expecting None make it easy to cheat?

rejected for not enforcing the use of a regexp

this has been fixed

Descriptions are generic, it is clear what tests expect when one sees the code.

FIxed some time ago

Don't make separated tests for None, or it's easily cheatable. You should create tests expecting None inside the same series. You should also shuffle the characters rather than just concatenating twice the input. Also, with this method, inputs expecting None will be statistically twice larger than pure random inputs; both should have the same average length.

The translation also needs to be updated to the latest version.

It makes no sense to have a 4kyu kata that may be solved by just using a built-in function. Learning to use built-ins is 8/7kyu level.