[Tancred]

It looks to me like the watched/unwatched status of a file is determined by it's path. 
I suggest determining it by a file's MD5 hash (or other hash), that way it would be irrespective of the file's location/name/URI and it would detect duplicates. 
Also moving watched files to another folder would retain the watched status.

This would likely completely break backwards compatibility, but I think it's more future proof and you could make it an experimental feature and keep the old code path for a couple of years.

[izprtxqkft]

i dont think you have thought this through

do you realize how long it takes to generate an md5 sum of a ~50GB file?

if not i suggest you find out, it would take days for my collection (and my collection is rather small by comparison)
then recheck all of them on every update?

just use an external solution like trakt.tv which tracks watched states based on content instead of file so no matter what file the movie is, trakt will know it is watched by the movie id


-----

edit:

for my collection you are talking about hashing ~45T on every library update just to keep track of every file

[Karellen]

Or you could use the WatchedList addon... https://forum.kodi.tv/showthread.php?tid=164652