I wrote before about this, but now I have some debug output.

(As it turns out I had to setq imap-debug before starting gnus to get
any output in the *imap-debug* buffer.  The startup only generated a
million lines -- about 55 MB on disk (once saved) and 110 MB of VM.)

I went offline to ensure no mail would be delivered and forwarded a
message (it happened to be that day's rfh post) to an address which
I knew would end up in an imap folder I had just emptied.  Then I
quit gnus and started up in a fresh emacs with imap-debug t.

Selecting the specified folder led to the:

 (assert (eq (char-after) ?\)) nil "In imap-parse-body 2")

from imap.el's (imap-parse-body).

A part of the debug looks like:

|||||| 7 -> imap-parse-body: 
||||||| 8 -> imap-parse-envelope: 
||||||| 8 <- imap-parse-envelope: ["Wed, 05 Dec 2007 19:30:02 -0800"
                                   "Selecting a child's name"
                                   (["tomkanpa" nil "tomkanpa" "aol.com"])
                                   nil nil nil nil nil nil nil]
||||||| 8 -> imap-parse-body: 
||||||| 8 <- imap-parse-body: nil

so the assert is in the sub-body which in this case is:

("text" "plain" NIL NIL NIL "7BIT" 714 25)

and not in the main rfc 822 body.

Also, the parse of the envelope gets nil for the sender, reply-to and
message-id fields.  I wonder whether that is where the parsing gets off
track?

The full debug (of just the one folder select) is about 36½K so I'm
attaching it gzip(1)ed (which interestingly is smaller than bzip2(1)ed).

-JimC