Wrote a python3 script (for a bit of practice with Python) to parse the CSV and the DVD file structure and match up images with identifiers in the existing records. I find only 29 matches. The script prepares the CSV file and a collection of images ready for import on the command line on the server, but I'm waiting for confirmation from AR that something isn't broken here; I'd expected more matches.

We've also discovered that the alternateIdentifiers field was not imported into the db when the CSV upload happened, because (I think) I misunderstood the documentation, or because (possibly) that import was a new feature not quite working when we tried it. I'm looking at the db structure so I can retrospectively try adding that, and I'm finding it's the usual mess of interrelated tables for a single field name and id (property_i18n, property, object, information_object). This can be done but it will have to be carefully tested.