If you watch the initial simulation (10 particles in each compartment) long enough, the two compartments do not really appear at equilibrium. Rather, the barrier moves in an erratic manner along the horizontal plane, reflecting a changing and uneven distribution of solvent molecules between compartments. If, however, you watch individual particles carefully, their movement is random. How can random behavior produce such obvious discrepancies in volume?
Because these simulations contain such a very, very small fraction of the actual number of solvent molecules present in actual solutions, their random movement across the membrane occasionally results in temporary volume fluctuations. Occasionally, a few particles, but a relatively large fraction of those present, diffuse out of one compartment and into the other without any returning the opposite direction, producing relatively large changes in volume. In much the same fashion, flipping coins may produce short runs of heads or tails (but over billions of tosses, the ratio of heads to tails rarely deviates from 1:1). Thus, the volume discrepancies here are artifacts of the small number sampling technique we're using to illustrate solvent behavior.