It should be possible to run succeeding tests faster.

By delaying an initial run for a shorter period of time and running more repeat tests, non-memoising functions would consistently time-out but correctly memoised functions could return faster. That would improve the user experience. ( I am not at all a fan of test suites that take ~10 seconds, and I just don't think it's necessary here. ) You're running tens of thousands of tests in some cases, so large numbers of tests should not phase you.

That said, even with that improvement, I don't know if I'd be very satisfied with the kata. I think the approach of my kata, with a curried function decorated at every level, is simpler and works just as well. I'm also not a big fan of variadic functions, that may play into it.

Hi,

My solution is raising TypeError, but still passes the whole test suite (I just see the error printeed in the console).

Note: the test feedback is rather misleading. Stuff like "don't change the object", considering your own solution, sounds like a bad joke... (basically, the kata is full of hiden specs, right now, and it's the best way to make the experience frustrating on the user's side).

Cheers