Some remarks on the history and future of averaging techniques in finite element error analysis