Exactly right. OC hits all the points. Fine granularity of failure, (re)warming caches, faster iteration by reducing cost of changes, etc. I too lived this. Al

Exactly right. OC hits all the points. Fine granularity of failure, (re)warming caches, faster iteration by reducing cost of changes, etc. I too lived this. Albeit with "poor man's Erlang" (aka Java). Our customers were hospitals, ERs, etc. Our stack could not go down. And it had to be correct. So sometimes that means manual human intervention. There's another critial distinction, missed by the "whole freaking docker-meets-kubernetes" herd: Our deployed systems were "pets". Whereas k8s is meant for "cattle". Different tools for different use cases.