Here is a paper on an analysis of the two-stream D/G/1 queue: main paper, (not needed on first reading App. A and B, App. C), Fig. 1-2, Fig. 3-4, Fig. 5-6.
This model tries to capture the situation where two VBR Continuous Media (e.g., video) streams are multiplexed. However, as most analytic investigations, it makes a number of simplifying assumptions and suffers from various constraints. The question then is to what extent this analytic model might be useful at answering more realistic questions. Thus one could compare answers one can get by use of the analytic results vs. simulation results of the more realistic setups.
Here is a number of ways the problem can be generalized or made more realistic: