I didn't disect it completely, or even execute it, but based on fist glance, I think it loads a DECIMAL(38,5) with a value that uses all the 33 digits before the decimal place, then multiplies this by 1. Because SQL Server hardcodes the result to have at least 6 decimal places and at most 38 total positions, only 32 positions before the decimal remain, and that's one too short for the current value.