We went through a similar arc. Kubernetes gave us a lot of theoretical upside, but for a small team with predictable workloads it mostly translated into operational drag: YAML sprawl, slow feedback loops, and time spent maintaining the platform instead of the product. Moving back to Docker Compose didn’t mean giving up discipline we still version configs, monitor aggressively, and automate deployments it just meant choosing a tool whose complexity matched our needs. The 60 hours saved isn’t surprising; it’s the compound effect of fewer abstractions, faster debugging, and less cognitive overhead. K8s is great when you actually need orchestration at scale, but it’s often adopted as a default rather than a requirement. This is a good reminder that “simpler” is sometimes the more senior engineering choice.