I found an interesting bug in the to_digest handler: when a message is not ascii only but has no encoding defined (which happens in the wild unfortunately), the mailbox will crash when adding the message on line 58 of handlers/to_digest.py.
The reason is that due to the mailman-specific email.message.Message modifications, some of the headers are unicode strings. As a consequence, when the mailbox.add code from the standard library serializes the email, it merges unicode strings with bytestrings, which result in an unicode string. If this string contains non-ascii characters, python will raise an UnicodeDecodeError doing so.
The solution for that is to call the as_string() method before adding to the mailbox, it will do the serialization properly.
I'm attaching the fix (one-liner) and a testcase to reproduce it (much larger).
Oops there was a typo in my patch, sorry about that. Here's the new one