Debugging service defects is a stressful job. You wake up in the middle of the night from an ops alarm (because system fail always happen in the middle of the night). You look at the stack trace and jump into your Clojure code base. Oh, that’s strange, this condition shouldn’t happen. You look into the database. Sure enough, the state of the task is really in this unlikely state. How did that occur? You load up your server logs to the timestamp where this exception happened. Oh, …