> All 4 are solved to our liking. Once we really need Kubernetes, we can do the switch. Indeed, but this still doesn't mean it's a sound technical decision to

> All 4 are solved to our liking. Once we really need Kubernetes, we can do the switch. Indeed, but this still doesn't mean it's a sound technical decision to build a bunch of custom tooling ontop of a dead and soon unsupported foundation that very few people use then realize too late that switching to a supported platform isn't as easy as you'd have hoped. I'm not comparing Swarm to Kubernetes, that's been done to death, I'm merely pointing out it's odd that you're spending time and energy building commodity features that already exist in k8s and are not core to your product. Good luck though!