N.B. note that Google performs some encoding conversions within the GMail backup from Takeout that permanently break all non-ascii characters when an encoding other than UTF-8 is used in the original mail (those characters are replaced by the 0xEFBFBD sequence, and since information in lost it guess it cannot be properly recovered).

Obviously I am not the only one who have been experiencing the issue, e.g. https://webapps.stackexchange.com/questions/71153/takeout-br...

The issue does not happen when doing backup with another tool e.g. mbsync.