Russ Olsen has a post titled Five Truths About Code Optimisation (via Bob Congdon). It's a good read, and it seems that point number 2, 'Rest assured, you don't know where the problem is' has struck a chord with a lot of people. That's because it's just so damned true.
It's easy to catch other people doing it, but despite learning the lesson over and again, I do it all the time myself. I've learned that a good corollary would be this:
keep your hypothesis to yourself until you've had a chance to prove yourself wrong
(By the by, Russ' blog is now added to my aggregator. Reading back through his old posts, it's an excellent resource).