When you have only two bodies gravitationally orbiting each other, their orbits are ellipses. Circular orbits are a very special case of elliptical orbits, and as such, no orbit in nature is perfectly circular (all of the orbits of the planets around the Sun, and of the moons around their planets, are actually elliptical, though the planets' orbits are typically close to circular). If you have more than two bodies involved in a gravitational interaction, then all bets are off unless you can approximate the situation as a two-body problem. Even as few as three gravitationally-interacting bodies only have nice orbits under certain special circumstances (for example, if two bodies are really heavy and one is really light, and even then, it depends on the initial conditions). Galaxy clusters contain many more than three gravitationally-interacting objects; in addition, these objects are not point masses, nor are they even rigid extended bodies. Rather, their mass is distributed in a rather loose disk or cloud, and it is easy for one galaxy to deform or even strip away part of another as it passes. This complicates the problem further, to the point where we usually resort to simulations to describe all but the most basic dynamics.