When the overflow is due to the max size a rational can hold, stop the reduction, otherwise some false results can appear. For instance: 1.0092^50*ln(1.0092) was computed to 0 due to this problem