Question about the real traffic pattern to Tor nodes

[Stone <snifferpro@xxxxxxxxx>]

     In the paper of Stop-and-Go mix, Kesdogan et al. assume that the
incoming traffic pattern can be approximated by a Poisson process. And
this assumption is also adopted by Diaz in the paper of "Towards
measuring Anonymity". But she questioned the assumption and in her
later paper "Comparison between two practical mix designs", she
concluded that the traffic pattern is highly unpredictable and is not
as commonly assumed in the literature. The conclusion is drawn by
examining the four months of real incoming traffic to a working
Mixmaster node.
     But I wonder that is the traffic pattern to Tor nodes or those in
low-latency system the same as that in high-latency anonymous system
such as Mixmaster. I think this result is helpful when analyzing the
properties of a mix( e.g. average time delayed before flushing, the
probability of n-1 attack,etc).  So Has anyone made efforts to prove
the traffic pattern in the Tor networks? Is the traffic pattern the
same as is examined in the low-latency network? Or can anyone help
collect some real input to examine it?  Thank you!