I think the disagreement here is that there's no evidence of this. Preview's error handling could very well be interpreting bad data and allowing the file to be opened. The question becomes, then, should Preview continue to propagate that bad data on save or should it try to correct it with the possibility that it corrupts just that data. If the PDF was not in-spec prior to Preview touching it but it is in-spec after Preview saves it, is it a good thing that Preview "fixed" the PDF file and made it "proper" or is it bad because it technically lost/corrupted data?

In other words, what is the "right" thing for a software to do in this case? Keep bad data and leave the file as "bad" or fix the issue to make a valid PDF and, as a side effect, remove the "bad" data?

Is it? I don’t see any accounts showing this to be a fact.