Just think this article as a training for bad managers... that believe that an 80% of code coverage means that tests must be good. In this case, having it achieve by random calls, shows without any doubt that they are wrong. That was the point. No the other way around. Sorry for the confussion Ken.