The idea is the don't cache the whole objects just the relevant parts This was recently added to k8s, see: https://github.com/kubernetes/kubernetes/pull/107507