At Oakland 2005, Murdoch and Danezis described an attack on the Tor anonymity service that recovers the nodes in a Tor circuit, but not the client. We observe that in a peer-to-peer anonymity scheme, the client is part of the circuit and thus the technique can be of greater significance in this setting. We experimentally validate this conclusion by showing that "circuit clogging" can identify client nodes using the MorphMix peer-to-peer anonymity protocol. We also propose and empirically validate the use of the Stochastic Fair Queueing discipline on outgoing connections as an efficient and low-cost mitigation technique.