This is something I would like to understand too, although my own focus is on late 1960s / early 1970s watches.
The case references on the insides of the casebacks always seem to be followed by a single digit iteration number in this period. There does seem to be some temporal sequentiality to it, i.e. the later a watch of a particular reference was made, the higher the number, but this doesn't always seem to be consistent (but maybe apparent inconsistencies are just due to swapped casebacks?).
I have seen nothing to suggest that there are any physical differences between cases with the same case references but different iteration numbers (except perhaps the addition of engraved designs on the back, with these being missing in most examples of iteration 1, but this could simply be coincidence of design change with iteration number progression).
Therefore I think your idea of the iteration number tracking how many examples of a particular reference were made is not a bad one. As the serial number is not specific to case reference, you can't use that to know how many watches of a given reference were produced in a given period. So it would make sense to have a case-reference-specific "counter".
But even assuming this is true, how many watches would be represented by a single digit change of iteration number? And would this figure be consistent across different references?