Sounds like it was written by someone who lacks good infrastructure to collect and search logs. E.g. it is very useful to collect all user actions and other ma

Sounds like it was written by someone who lacks good infrastructure to collect and search logs. E.g. it is very useful to collect all user actions and other major actions. So later when user face some problem and sends support a message you know what was going on. You can send user a message that his issue was fixed instead of asking the user for a reproduction steps. The aggregate numbers poorly tell you a story what has happened to this particular user and at scale there are bugs that just some ppl will hit. You can process logs and produce the metrics out of them, but you can't do the opposite. Saying logs are just human readable actionable errors seems backwards to me.