If somebody sends non-spam to an invalid address I'm sure they'd like to know, but spam should get silently dropped as it does now.[/QUOTE]
Sure, spam filtering will cut down the traffic, but no spam filters are 100% effective. Plus, if you are not getting any spam, you probably are getting "false positives" (i.e. legitimate email flagged as spam) which can be worse.
And unless you are syncing the SpamAssassin Bayes databases between the two boxes, the backup MX will likely be less effective at filtering spam than the primary, since the primary will get more spam (autolearn) and the users there will be using the Junk button on the web UI.
Our system gets about 6,000-10,000 emails per day. And this is after we filter out traffic at the firewall via RBLs, anti-virus, anti-spam, and anti-phishing (the firewall drops about 50,000 emails per day via RBL blocks). Even with all that filtering at the firewall, about 10% of the remaining traffic is filtered out by Zimbra as spam/virus emails.
So, for us, any amount of backscatter would be a big deal, hence, my tenacity at trying to make you a convert to recipient verification!
All the best,
Mark