Reviewing this again, it seems to me that there are two separate issues reported here: (1) set_payload on an existing MIMEText object no longer encodes (but it has now been a long time since it changed). (2) the functions in the encodings module, given an already encoded message, double encode.

(1) is now set in stone.  That is, it is documented as working this way implicitly if you read the set_payload and set_charset docs and has been working that way for a while now.  An explicit note should be added to the MIMEText docs, with a workaround.)

(2) could be fixed, I think, since it is unlikely that anyone would be depending on such behavior.