Thanks, @hobovsky, your advice helped! Seems like I underestimated maps in Java =)

Thank you, hobovsky, I just posted it there.
As for how far it gets, I see it goes from 1x1 to 100x100 maze size several time. I see that mine solution runs 1290 times, meaning that it goes 12 full cycles.

Ah ok, I get it now. Yes, CW runner kills tested process if it prints too much to the console. But still you can investigate how far you get by printing something smaller, like just he size of the maze, or size of your history, right? This should not exceed the buffer limit. Another gotcha is that when process is killed on timeout, output which is not flushed can be lost, so be sure that stdout is flushed after each print.
If you are still stuck, you can post your code for others to check (remember about markdown formatting and spoiler flag, or, even better, ask for some advice on kata solving gitter chanel.

You missed the beggining of the sentence. I get buffer overload only when I try to print out each maze. So after it prints bunch of 100x100 mazes it says that buffer is limited to 1.5Mb and it is overloaded or something like that. When I don't print each maze, I get time out error. I was trying to see how far I get when I try to print maze, so I can understand if I timeout on 10x10/50x50/100x100 maze.
Anyway, thanks for the reply! So history is not an issue, I got it. It actually makes it even more confusing now =))))

I get buffer overload error or something, so it seems like my algorithm is fairly good...

Your implementation crashes, but otherwise it has to be really good indeed :D

Jokes aside, you need to store some kind of history to solve the problem, so that's ok. However, tests seem to be kinda challenging, because my solution takes ~7.5s out of 12s allowed so you do not have too much time to waste.

Getting timeout(language - Java). When I print out maze itself, I see 100x100 maze and I get buffer overload error or something, so it seems like my algorithm is fairly good... Can anyone say if it's possible to solve this one without using history to store taken steps? My algorithm uses pretty much minimum amount of steps and only required steps to determine if maze is solvable, but I store every step I take and check if step I'm about to take was already used. I wonder if that's why I'm getting timed out?

I got same issue: array lengths differed, expected.length=0 actual.length=20
Seems weird, since input is: 15 17 0 29 20 27 33 32 25 0 34 3 31 13 9 8 1 5 8 27 31 29 13 19 25 17 14 25 7 19 19 29 7 0 17

Can anyone help, please?