Even if equality was by memory location in the case where the tuple contains pointers, one would still have to work out how equality would work for tuples of no

Even if equality was by memory location in the case where the tuple contains pointers, one would still have to work out how equality would work for tuples of non-pointer types. For example, would (NaN, NaN) match (A, A)? If the implementation makes use of ==, then it wouldn't, and if it uses bitwise equality it would.