Metrics are great. In general, I subscribe to "don't log successes" (with the caveats such as log events tied to billing or that get passed as messages to other

Metrics are great. In general, I subscribe to "don't log successes" (with the caveats such as log events tied to billing or that get passed as messages to other services). But what if your problem only occurs at scale? You won't be able to replicate that you in your dev env. Error logs and event logs make sense in many (most?) services.