I agree, but for a slightly different reason. Treating currency (with decimal digits) as floating point is a sure way to end up with rounding errors. Try to explain that to a manager who hasn't touched math since high school. That is why I really appreciate Oracle's NUMBER data type.
I think the original post was in reaction to Oracle's statement that it couldn't spread the F.P. operations across multiple CPU's. My guess is that the overhead to move operations to other CPUs would outweigh the benefit, but that is just a simple guess.