Bro, this is absolute gold! I was literally struggling with this exact same thing last week. Have you tried applying this same approach to larger projects as well?
Thanks! Yes — Nyx runs in production on AKS and is fully tested on EKS as well. The approach scales because collection is scoped per-namespace rather than per-node: the DaemonSet sits on every node, but you choose which namespaces get monitored, so a 200-node cluster with 5 monitored namespaces costs about the same as a small one. The lab runs on kind, but it's the same enforcement path we run on real clusters.