It still says in the task description that bases above 21 won't be tested. Although for my python solution it is timing out being tested against base 35.

either way, the suggestion won't be implemented, as it would invalidate too many results on an approved kata

Go ahead ;)

Shall we redo the tests then or would you like to?

Current top solutions can all handle up to around n=30, give or take and all are unoptimised.

for me, it was not clear at all from the description, which sequences to include and which to discard

Yes, for a 4 kyu, let's push this somewhat.

This is now fixed.

Only invalidated solutions were the ones which tried to use a correct prime checker and then append the wrong results as special cases.

Okay I have approved your fork with some modifications. If this invalidates solutions then it's on me.

You shouldn't end up invalidating any solutions if you add the import to the setup as well as preloaded.

It'll be disgusting, yes, but we haven't been given much choice.

I've made a fork fixing this issue by using an exterenal library.

Honestly, I believe it would be better to explicitly import it in the initial solution and not mention it in the description, as that would make it easier to translate. However, as to not invalidate any solutions, I just messed with the preloaded section.

not enough random tests.