Is there any evidence in this paper or elsewhere that this is causation rather than correlation?

For example, it seems logical to me that people with worse health and failing mental faculties will already be feeling more motivation to retire earlier, as opposed to very healthy people who will keep on working forever. That would be pure correlation

Yes - very roughly, they find that regions with lots of layoffs show an increase in subsequent cognitive decline:

https://www.nber.org/system/files/working_papers/w35117/w351...

https://en.wikipedia.org/wiki/Instrumental_variable

Not perfect, but less vulnerable to the correlation issue you mentioned.