This year, I'm going to share some amusing (or scary...or both) anecdotes from the conference. In some ways, I think these are almost more important than some of the substantive information, because there's always that thing about learning from history.
In the talk on performance regression, Marcel mentioned that someone had turned off the combiner for Yahoo! Profiles. None of the continuous deployment tests detected this because there were no performance tests. Of course, the RUM monitoring found this in production, but this is the sort of thing that should never happen. Has something like this happened to you?