Hi Olly,
As per my understanding to what you are concerned about:
In the exported "existing.xlsv" file(its format should be xlsv instead of csv, sorry for incorrect typing of my first reply), we can find that there are 3 hidden columns if we pull the column D to right, and the first column corresponds to record GUID.

Therefore, it indicates that this re-import job will only update existing records, every records will be match correctly, it wouldn't fail if we only modify columns(fields) with correct data format.
Please let me know whether if I misunderstand the problem you are caring.
Regards,
Clofly