I'm not sure your issue is so much an iptables problem as a routing problem - if your primary connection fails, and you get incoming traffic through the second router, your SME probably "replies" to the primary router anyway (since it's the default gateway), which then doesn't know what to do with the packets. To fix this with iptables, I think you would either need a rule to change the routing for any traffic that had been passed through 10.0.0.3 - which I've never seen, or you would have to configure a second WAN ip on your SME designed to talk to the secondary router.
I can see 4 ways to solve the problem:
1) Modify your primary router to use the secondary router as its gateway if its internet connection is down (this may be possible on some routers, but I don't know any specifics about how to do it).
2) Write a script to run on your SME that verifies the primary internet connection, then changes the default gateway in the event of a failure. That is, try to connect to some off-site host, and change the default gateway on failure.
Note that in both of the above situations, your backup MX record will fail for all connections unless/until your SME server "fails over" to the backup internet connection.
3) (The easiest solution) Purchase and install a
Xincom (or other) dual WAN router between your two ADSL connections and your SME server (note that this solution becomes tricky-but-not-impossible if you're using any sort of VoIP on your LAN)
4) Setup another SME on the secondary internet connection, configured to use the first SME as an "Internal" mail server. The proxying capabilities effectively "re source" the incoming email streams, so email works OK. (While you're at it, make the second SME an Affa backup of the first SME...)
This last option still leaves you with a problem sending emails and browsing the internet if your primary WAN fails, but at least email will continue to come in...
To resolve these last two problems you would still need a script (as in option 2 above) or a manual procedure that either changes the default gateway, or configures "upstream" SMTP and HTTP proxies to pass through the secondary SME server -- but at least you could SSH into the secondary SME, then get to the Primary to make the required changes from off-site...