Could we generate UUIDs for the imported mail? That would be incredibly
valuable in my analysis _anyway_.

Hmm yes I could create the users in Mailman3 even if they never sent an email in the new system. But I may have a better idea, here's what I could do:
- get the old emails from Mailman2's mboxes and the new emails from Mailman3
- strip the non-text content
- replace email addresses with hashes to hide them
- generate mboxes or maildirs

Since your tool works with mboxes (maybe maildirs?) that would give you something you could import and run stats on. Would it work for you? Am I missing something security-wise or privacy-wise?

Aurélien