That's a really wonderful puzzle. It took Stockfish on my MacBook Pro quite some time of crunching until it displayed correctly that it's a mate in 17. At first glance, I thought that the unexpected move 1... Kh8 could maybe require a solution beyond 17 moves (because Stockfish was unable for quite some time to display a definite mating sequence of any length), but after some more searching Stockfish could confirm that 1... Kh8 leads to defeat in less than 17 moves as well.
The puzzle is great, because it allows for several different branches which all lead to mate in exactly 17 moves. I guess that's why it takes Stockfish quite some time to search through all the possible trees.
Here is another interesting puzzle which is currently totally beyond Stockfish's reach. It's a mate in 36 with a single bishop from the Lomonosov 7 piece tablebases:
en.lichess.org/study/RDaWF37dIt's also highly interesting, as every slight inaccuracy in attack or defense leads often immediately to a draw or a win for the other side ... or, alternatively, to longer or shorter mating sequences.