Ignoring castling rights and en passant state in zobrists hashing

[koedem]

Not sure if I missed this in the thread, but I want to point out that different en passant states might happen not just after repeating moves, but also for different move orders, consider simply:
A 1. e4 e5 2. d4 exd4 3. c4
B 1. c4 e5 2. d4 exd4 3. e4
Which are the same apart from different en passant rights, no repetition or anything involved. And it seems easy to believe that something like this might break in a real position. Does it cost a lot of Elo? Probably not, it might lose a bit and gain a bit due to less overhead. But it seems unsatisfactory to me at least, to occasionally break down in simple scenarios like the one above.

[Henk]

Currently I don't trust anything stored in transposition table. So I use zobristkey only for three fold rep check etc.
But then zobristkey must still be accurate.