Comment 5 for bug 1859011

To clarify some points:

1. the bounce mail is genuine. My logs show that a ~5MB mail was send to the
list by the owner of the list, it was distributed to the members of the list,
and caused some bounce mails by some recipients. The mail that caused the
problem was normal non delivery notification with the characteristic that
the complete original mail was included as message/rfc822.

2. I don't know if the size of the bounced mail has an impact on the number
of duplications, but the size combined with the duplications resulted
in the bounce process requesting using ~20 GB of ram

3. My logs indicate some cases with >2000 processed bounces that didn't crash
the BounceRunner, most likely there was also a duplication, but the mail
was smaller

At the moment I try to create a striped down version that I can share
that reproduces the problem