Fixed points of monotonic and (weakly) scalable neural networks