Makes me wonder how many more papers out there have hard-to-pin-down errors like that

And how useful potentially AI could be to spot those (even if retrospectively)