Very nice test position! My chess programme (Milady) does not solve it, though at depth 16 the evaluation decreases a little. Probably it overestimates the white castling destroyed; it also could overestimates the weakness when 1...g6 is played. As a result it prefers the sacrifice! It is a weak engine but I estimate it at 2100 nevertheless. I could easily add a trapped piece term in the evaluation (I remember older versions may have such term) but this additional term has a cost.
Thanks again!
Yves

__________________
http://milady-chess.blogspot.fr/