there are a couple of theoretical approaches at a network level, but neither of them are going to yield any form of success in practice. you could do this via multilayered tunneling (which has always been a problem - i don't know of anyone who's implemented it with 100% success). however, each hop would be indifferent/inactive to the action - they would simply be a virtual gateway to the next hop. so in effect you would only be tracerouting the end target via a series of encapsulated gateways.

the other route would be through source routing options - which would allow you to specify a next-hop address - but the port/transport(less) information would remain static.

to do either would require that functionality exist and be available to you at each router/hop in use. and you'd be lucky to find any public router that supports source routing, period. but there are a number of routers that support gre by default. however they are likely to be bound from within a protocol that is configured to require some form of authentication (and optional encryption)...ie, pptp. and even then i'd say that the ability to tunnel within a tunnel (which is generally considered impossible based on lack of implementation - although completely feasible on paper) would be significantly limited.

the only other option i see would exist at a service/application level, whereby you are simply chaining proxies. essentially the same as the above, but implemented at a completely different layer, and generally specificly designed for a particular application layer protocol.

...but that's why they play the game. give it a shot, see where you get with it. record your findings and get back to us with an update.