Very interesting explanation.
As regards Python floats, they are native floats (64 bits long in IEEE 754 format), not arbitrary precision floats. Only integers are in arbitrary precision. So, I think that the rounding errors are also present in the Python program.
