Through the GMail API, can I assume message_id is unique amongst all the messages I index, or would I have to rely on the (message_id, thread_id) tuple?
Just the message_id alone is unique across the inbox (for that single user).
c.f. https://developers.google.com/gmail/imap_extensions#access_to_the_gmail_unique_message_id_x-gm-msgid