You want to simulate an infinitely large chaotic process in a finite amount of memory??
The small objects: are they anchored? If not, how buoyant are they compared to the estimated lifting ability of the fluid? If they aren't anchored then there may be gravitational or electrostatic attractions between the objects, or the wave motion might happen to force them in to proximity: once in proximity, the viscosity of the fluid could act to "drag" them with respect to each other, resulting in one "climbing" on the other. Such effects would, it seem to me, require quite a fine-grained simulation, but the finite amount of memory you have for the infinite simulation suggests you will not be able to provide that granularity. Or are you planning on fine-scale simulation within a certain distance of the objects and mere statistical non-chaotic calculations beyond that range?
What model of the geometry of the Universe are you planning to use? If you model it according to the current models of the curvature of the Universe, then your infinite cylinder might have to be a closed torus -- but a closed torus has different wave propagation modes than a cylinder does. If you instead model the Universe as infinite and linear and as containing an infinite amount of matter (an assumption needed in order to be able to fill the infinite cylinder with fluid), then I suspect that it could be shown that the probability is 1 that some portion of the cylinder will pass close enough to the gravitational influence of a massive object (star, black hole, significant planet) that you will need to include gravitational influence in that area of the cylinder.
