I am not an expert (and don’t have any idea what I am talking about), but wouldn’t this be captured the same way other future positional advantages are when evaluating the next “level” of possible board positions? Ie, the fact that a piece can later give check, or castle, or anything else if you first move it here is a function of that next move’s resulting position score.

This network is just to signal to the final search algorithm how good the board looks in a given spot on tree of possible moves.