Hi Tanis,
I may have been premature in sharing that tip - I apologize, but it appears it may no longer work!
I was basing my suggestion off of this ticket:
Back when we were working in 2.2, a developer and I discussed wanting a feature that would allow you to import translations, and filed this ticket. The developer I was working with at the time ended up marking the ticket as invalid because he discovered that this had already been added to the code, but never documented. I haven't yet documented it formally myself because I hadn't had the opportunity to test it out and discover exactly how it works, and its limitations, but it's something I've been meaning to investigate for a while.
However, based on your reply, I did just perform a test, which also failed. I tried it both from the command-line and via the user interface, and in both cases, new records were made for all rows, and there were no warnings or errors when I used the same legacyID values for multiple rows.
The CSV import code has undergone a lot of changes in both the 2.3 and 2.4 releases, and it's possible this legacy functionality was removed or overwritten at some point. Previously, if a CSV had the same legacyID value in more than 1 row and the culture values were the same, the import would throw an error and halt. As per the ticket linked above, if they had the same legacyID values but different culture values, then they should have been imported as translations.
I am going to update the ticket linked above as new, and move it to the wish-list project, so we can review the functionality and hopefully include it in a future release, though it may require community sponsorship for us to be able to do so. For my own curiousity, I'm also going to ask around about this - unfortunately the original developer is no longer with us, but hopefully someone can point me to where this code is in older releases, and why / how it was removed.
In the meantime, sorry for pointing you in the wrong direction! I regret to say that without this, I don't currently have any other suggestions for bulk translation in AtoM at present.