For something as extreme as a blank method, it's unlikely to be useful, I agree. For something subtle in modifying existing code, "did that actually change the thing I thought" might be a question worth answering. I guess my best (somewhat Devil's advocate) argument for doing it all the time is that maybe it's cheap enough that doing it in the useless cases is less expensive than every time figuring out whether you should plus the cost of the false negatives.