Shouldn't the most immediate response be, "why aren't you memoizing"? ;-)

So I think the idea is that users be able to suggest test cases breaking some existing solutions and probably get rewarded for it.
As for suggesting anything, you can write that in a comment; without a reward, though. As for "Playing with digits", the kata, like a lot of others and unlike competitive programming problems, is not defined strictly enough. It isn't clear what the intended difficulty was meant to be and thus whether big number tests should be there.

Yes, I can. But, will be "clever"/"best practices" rating recalculated in this case (if somebody hack solution)? Mistakenly solution cannot be in the top.

I'm still using Windows XP quite often and the only problem with any web site I can recall is limited availability of installed video codecs for browsers to use, so it's probably not about Windows XP.

@vudulin same to me, are you using Windows XP? remove s from https should fix the issue ;-)

@vudulin This is a weird error that seems to be happening perhaps based off of specific network conditions. You are the 2nd person to report this. You can try changing the page to not use https which should function as a workaround until the issue can be fully resolved.

Would it be possible to add a configuration option for random language in the katas?

You can fork any solution and add tests, change code, etc. Is that what you were looking for?

What's about adding possibility for hacking another persons' solutions with own test cases? (like in Codeforces system). I often see a lot of submits, that obtains unreasoned solves working with existing tests. But these realisations probably will be crashed with other inputs... and it leads to disastrous results in the future during working with real code on real projects.

For example, let's see on the popular task "Playing with digits" (https://www.codewars.com/kata/playing-with-digits/solutions/csharp/all/best_practice ). That's enough to take two big numbers for input, that crash 95% of solutions with TimeLimit. Also a lot of resolves don't have a processing of illegal input params. As for me, critical input values should be a part of tests. And in such situations (when task creator didn't provide anything), other participants should be able to fix the omission.