I beg to disagree on that. Where in the work flow the additional image(s) enter is immaterial. To be clear: I do prefer some kind of early stage merging, possibly in-camera, as this tends to reduce the processing workload overhead. But this is from a perspective of being short on time, not based on principle. As I already exemplified, 'layering' can be done in so many different ways and I used the various techniques a lot in my film days too. Way before anyone dreamt of PS.
We haven't even started to discuss beam splitters and mirrors. Superimposing frames is in photographic terms really old hat.