As I understand it, the answer to the "arch/spandrel" situation is as follows:  The distinction between an arch and a spandrel is external to the process whereby arches and spandrels are constructed.  Similarly, the distinction between coextensive traits is made externally to the process of natural selection.  The fact that one can distinguish between coextensive traits does not imply that any laws or processes make the same distinction when speciation occurs.  So I do not see a problem with Darwin here.