I can do sender/receiver unlinkability to twitter/*chan no problem. Cf. Matthias Bauer, "New Covert Channels in HTTP", WPES 2003. The more I think about it, the more I think that trusting the sender ends up being a social problem, but I'm going to keep chewing on it.

Also got pointed earlier today at Heydt-Benjamin/Serjantov/Defend, "Nonesuch: a Mix Network with Sender Unobservability", WPES 2006, which I need to read more thoroughly but also looks promising.