That's a really good idea, do you know if there are any tools out there for doing this? I'm thinking for something basic you could make it change random variables or numbers in the code base and run the tests after each change and report if they failed.
Fuzz testing involves an automated process commenting out each line of your code one-by-one and running your unit tests to see what lines are actually tested.
I wonder how well this would work for those Inuit guys with such and such millions of LOC ;). "Wait, what Big-O complexity does this code have again you say?"
