I have updated your patch, changing wording of some of the documentation, and applying style tweaks to the code.  I have also done the refactoring of the _set_xxx_state method that I suggested.  Seemed easier to show you what I meant in code rather than try to explain it.

You should use Reitveld's 'patch diff' facility to look at the changes that I made to your patch. 

Things that remain to be done:

I'm getting warnings when I run the tests.  These should be either suppressed or checked for (assertWarns).  (Or fixed, in the case of assertEquals :)

In the process_message docs it says that it should use RFC 821 format for the return, but in the new process method it says RFC 6531.  This makes sense for the new method, but is there a more recent RFC the old method should be referring to?  Or is there an open issue about more modern return codes for smtpd?  I seem to remember something, but don't have time to look now.

There is an issue with the reset of the maximum command length, but that can be dealt with in the 'duplicate HELO/EHLO' issue.

process_smtputf8_message in DebuggingServer should be receiving bytes, and so should error when it tries to split via '\n'.  Presumably this means there is a missing test as well, and also the same issue if decode_data is False when using DebuggingServer.  (Unless I"m missing something; I didn't try to test it.)

Different issue, but have you given any thought to what it would take to make PureProxy support SMTPUTF8?