SPF is an optional system that somewhat acts like you propose. SPF is an add-on to the DNS system, wherein email administrators say "if you get an email from my domain, then verify that it came from one of the following servers. If it didn't, it's a fake ... treat accordingly". It's not perfect, and it's not ubiquitous yet, but I think it is one component of a successful anti-spam strategy.

Another component that interest me is greylisting which takes advantage of a mechanism already in place in the SMTP protocols. It, too, by itself, is not a panacea, but in conjunction with other methods may prove to be very effective with a small amount of effort on the administrator's part.

Between these two methods, one might be able to weed out some percentage of the "laziest" spam, meaning your Bayesian filters and other layers of defense have fewer messages to process. SFP and greylisting can be the bouncers at the door, merely checking for valid ID's. Many will be rejected at the door. Those who are able to get past this check and into the club will be frisked by bayesian filters and whatnot...