Fractional part
Defects where code assumes a floating-point value carries a meaningful fractional part that is in fact absent. Integer division feeding a float, an earlier truncation, or a value that was always whole leaves nothing after the decimal point, so logic that depends on that fraction — a remainder, a sub-unit quantity, an interpolation weight — operates on zero and quietly produces the wrong answer.