Dead email accounts cause launchpad to mailbomb and crush subscriber inboxes

I guess at a bare minimum launchpad should have some rate control on the number of comments that can be received consecutively by email to a single bug report in a one hour period. The potential for abuse is also there because you could sign up someone else's account then set this problem off...

Note that the bug mail sent out by Launchpad set the envelope sender to a separate email address (<email address hidden>), and also set the Errors-to: header.

The problem account appears to be https://launchpad.net/~nomorespam (which has its email addresses hidden -- will need admins to check this) who is a direct subscriber of bug 71657. Looking at the ubuntu-bugs mailing list archive, I find the following message for that bug:
    https://lists.ubuntu.com/archives/ubuntu-bugs/2006-November/343736.html

Now that message claims to be from an @gmail.com address, but the errors come from @shell.visi.com, so it is possible that there is a broken forwarder somewhere.

Why were comments on a duplicate bug report being sent to subscribers of the original bug report? That doesn't seem like a good idea.

Matthew:
I guess it is so that any comments are seen by the widest array of related people - e.g. if someone doesn't know they should post to the original bug their comments will still be seen. It also means that if someone says "hang on this bug isn't a duplicate" again a larger number of people will see that comment.

On Sun, May 13, 2007 at 10:36:58AM -0000, James Henstridge wrote:
> Now that message claims to be from an @gmail.com address, but the errors
> come from @shell.visi.com, so it is possible that there is a broken
> forwarder somewhere.

I've looked at the e-mail that was received by Launchpad, and it's
indeed a broken forwarder; it doesn't send any obvious headers we can
look for to determine that it's not a valid comment. I'm looking into
how to handle this better.

My current bug page branch includes a hint, under the comment field in duplicate bug pages, that any further comments should be confined to the issue of whether it really is a duplicate. I think that's a better solution than broadcasting comments from all duplicates to subscribers to the original.

Help... (Bug #114943 )

Setting to critical since this happened again. I have a fix ready, just need to get it pushed out to the production servers.

Fixed in RF 4240.