I recall reading that giving different groups the specs like that and writing different code doesnt actually work well. They tend to write the same bugs. In this case it may he that they got to choose the best of the 3 versions, so it may not have been a waste.

Formal unit testing is another idea that turns out to be relatively low value but is widely regarded as a best practice.

IIRC NASA studied both of these, but I dont have a reference.