Answer from cs61c-ao (Roger Tseng 16507336) for Question 1
someone could exploit this by storing say an instruction into an int and then when memory loads it, it won't know what to do with it.  or if an int was given a float then the precisions would be off.  one possible technique could be to type check by letting some bits distinguish the types.  however this will use up much sought after register space.