Randy Layman wrote:
>
> The problem is that neither are directly subscribed to the list -
> they are the result of mail forwards (I've already asked Craig about this).
> The only way to find the bad addresses would be to send an email to every
> address separately and determine which outgoing message produced the
> failure.
And it IS possible to do just that without interfering with normal
operation; just add a random character (to a header field which get
replied) for each normal ML message which is different for each
recipient.
--
- Torgeir