This paper is weirdly similar to one I wrote for #NeurIPS's "I Can't Believe It's Not Better!" Workshop. Had I not been constrained by the structure and rules of the workshop, I would have pretty much written this exact paper. The title is even quite close to "Pitfalls in Machine Learning Research: Reexamining the Development Cycle."
http://proceedings.mlr.press/v137/biderman20a/biderman20a.pd...