Can you elaborate on the Kubernetes bit

I'm not GP, but I imagine it has to do with efficiently scheduling pods onto nodes to optimally support workloads, some of which have a resource affinity (CPU, MEM, Disk) that can only be supported by particular nodes. In this analogy the affinity would be a strong preference for isle and window seats or sitting with family. It's easier to have the pods sort themselves according to preference than to write a daemon to do it.