My experience is that we validate our assumptions when debugging far less than we should. Most of the time your assumption can be tested very trivially, while you can waste a huge amount of time if it's wrong.