That's just what I wanted to say. I call it "debugging by superstition." I think there's a fine line between this and a useful hunch, but there is a difference. Superstitious debugging is rapid-fire, change-a-line-and-try-again. A hunch extrapolates from the evidence and your experience, and it leads to an experiment to prove or disprove it. In the sysadmin world, Rachel by the Bay has some great stories about following hunches.