Postmortems are how we publish negative reliability results.

I spent time this week at a CS systems workshop. There was a lot of discussion about the importance of negative results in CS systems, and how they don't get published.