I believe that this is a very viable approach. I am currently doing research on intelligent intrusion detection systems (not based on traffic analysis), and while SPAM and IDS don't seem awfully related, both traditional traffic-based IDSs and STP utilize data traffic analysis methods to identify potential problems. That being said, I think that the traffic analysis should be used only in combination with existing spam control and heuristics; it's a complex and multi-faceted problem and thus requires several fronts to combat it.