Again you are not right about that. In practice in the vast majority of the cases¹ the editorial board (for journals) or the program committee (for conferences)

Again you are not right about that. In practice in the vast majority of the cases¹ the editorial board (for journals) or the program committee (for conferences) do that. And these people are researchers, and they're not paid by the publishers. At this stage the publisher does not intervene. ¹ In some other cases such as with Nature, what you said is true. But the point is that the model you are describing is far from being the only one, and there is no evidences that it is better. In computer science for example, the most prestigious venues are conferences not journals, and in that cases the publishers does nothing until the very end of the process, where its work often consists in putting PDFs online and that's it.