We wrote our own CSV parser to get fast import performance, and we do occasionally encounter novel encodings and weirdness. I sent you an e-mail to get more details.