Thanks for the tips. I don't think e-mail debugging would work for me
because not all e-mails are failing; only some. In reviewing the code
base, I didn't see any substantial changes that happened around where
MessageSendErrors are generated.

In deciphering the string of the message, it appears there's only one
place in the code where that exception could have originated, namely,
mailer.py:287. So, against 1.4.21, I've added the following patch and
installed it:

diff -r d6e9f95cc30e roundup/mailer.py
--- a/roundup/mailer.py Fri Dec 21 15:06:23 2012 +0100
+++ b/roundup/mailer.py Tue Jan 01 17:09:51 2013 -0500
@@ -284,6 +284,9 @@
                 raise MessageSendError("Error: couldn't send email: "
                                        "mailhost %s"%value)
             except smtplib.SMTPException, msg:
+                with open(r'c:\inetpub\Adams Row
Tracker\exception.txt', 'a') as file:
+                    traceback.print_exc(file=file)
+                    file.write('\n')
                 raise MessageSendError("Error: couldn't send email:
%s"%msg)

 class SMTPConnection(smtplib.SMTP):

My hope is that diff will provide access to the underlying error that's
masked by the MessageSendError if it occurs again. I'll report back if I
find something.