Detecting language please wait for.......
The use-case would be WS projects where the pre-translation is done against usual translation memories, AND translation memories that contain raw machine translations. The idea is that translators can see the translations coming from the "machine translation memory" marked as AT and not as full matches.
I have responded to this question in a support case to your colleague Giuseppe today. I have also created an article about this that I believe provides the answer:
I hope this helps you,
Many thanks, Caterina, it was really helpful.