[IneffablePigeon]

> If your app is constrained by CPU or RAM, then the business metrics will reflect that, and then you can turn on collection of those metrics to identify the problem.

…but why incur that round trip on my feedback loop? Having those metrics on doesn’t cost me much.

This feels potentially like the perspective of a large organisation with both mature monitoring systems and quite steady state user base activity (through scale). When I have a customer who had an issue yesterday because they had an unusual workload that won’t be repeated often, I can’t afford not to have had the basic metrics turned on, in case they point us in the right direction.