I found this article written by Jason Barile via James Avery (who found it on Rob Caron's blog).

http://blogs.msdn.com/jasonba/archive/2005/07/14/438812.aspx

“In the end, you need to reproduce the problem and understand the core issue before you can have confidence in the fix.  It's important to remember that you can always prove something doesn't work, but it's really hard and often impossible to really prove something does work.”

I'm always surprised at the number of times people I work with blindly try to fix a bug, even though they can't (or won't even try) to reproduce it.  It's worse when they won't try to get to the core issue, but will simply fix the symptoms they see.