I know on the 16C5x parts the code protect bit erases when the rest of
     the chip is erased.  I _think_ that it probably takes more exposure -
     I haven't done any real experimentation to determine if this is true.

     As to making the code protect bit un-erasable, I don't think that was
     intentional.  Even the wonderful chip designers at Microchip sometimes
     screw up.