It certainly sounds like a good idea. I guess the real question is: if the spamming software in use is given a temporary failure when trying to send to the secondary MX server, will it immediately try to send to the primary server? If so, your idea will have no impact. If you could set this up as a test, it would be most interesting to hear the results.
It's possible, but it seems a lot of spamming software is completely stateless. It doesn't retry at all, which is why greylisting is effective.
In my network, only 8% of the hosts ever given a 4xx by my greylist attempt to deliver again within 5 days. My Greylist is very short (1 minute), so pretty much any retry at all will get past it. Given that a lot of spam can't even leap over this short hurdle, I doubt they try the primary MX unless the secondary is completely unavailable.
I'd definitely love to hear what the results are. Theoreticaly they should be good.
