There's some real truth in this link - we do tend to underestimate bug time, and these are some real reasons why this happens. However, good time estimates are very useful to the people requesting changes - after all, they only have so many hours to "spend" and they do a much better job of budgeting those hours if they had an idea of the cost for each feature or fix. It gives them a sense of perspective.
So I'm curious how other folks handle time estimates? Are issues estimated on the way in? Does anyone log the actual time, for calibration? Does it seem to help? What seems to work?
(yes, I'm fishing here, any useful experience would be appreciated)