This is intended to help protect users against #1671. It doesn't resolve the issue, but I think its a good stop-gap solution for 0.5. A "full" fix would be very involved (and maybe not worth the added complexity).