We’ve just been trying to do this with our hosts’ help, can’t see any obvious appearance of where they’re being created using the access log, although we’re not entirely sure we understood the log completely!
I think I may have figured something else out though that may be of use; they all contain tags that the user would have added. Selecting one relatively obscure tag that I believe would have only been entered for one post, does indeed filter to only one of the problematic posts.
What’s more, these posts all have the date of publishing as either 2012 or 2013. The earliest, and unaffected post, after this is from 2014.
So is it likely this isn’t to do with a hacking?