@B4rrCH, already figured it out thanks. Trick was that test suite runs 1x1 up to 100x100 random tests five times not just once.
What I did initally (and probably you too) was short-circuit to a wrong answer on first occurrance of 100x100 in which case it would halt in mere seconds, not 12 sec timeout. However short-circuiting on 2nd or 3rd occurrance of 100x100 makes it clear that randoms run multiple times.

Tho there are absolutely no differences between 99x99 and 100x100 on the tests side. :/

and a reasonnable amount of solutions already exists... So that should be on your side, even if hard to track (local executions aren't "trustful", since you bypass some kind of interactions. Like: would you have a global data structure that doesn't get updated/cleared correctly at some point? You won't ever see that on just one test, run locally)

Same issue, codewars passes randoms up to 99x99 in less than 4s and then hangs on 100 x 100 every time.
I dumped the input it completes with correct result in matter of milliseconds locally. Must be some limitations of the codewars sandbox?

UPD: OK it times out because random tests run multiple times (not clear how many times exactly) not because of any particularly hard puzzle. Should description be more obvious about perfromance expectations?

I don't do C#, sorry. about the time out, the idea it's that if you print enough from the right place, you'll get a buffet limit error befits the time out. Then you get some feedback

print the inputs and print stuff into everyone of your while loops, to make sure that your videdoesn't get stuck into an infinite loop in some cases.