As noted by @MikeG in this comment, the Universe Sandbox 2 FAQ addresses the question, but not in a satisfying or helpful way:
Does it account for relativity?
No, the physics in Universe Sandbox² is currently only Newtonian.
Why? The short answer is that you need a supercomputer to accurately simulate general relativity. (emphasis added)
Jenn, astrophysicist and Universe Sandbox² developer, explains more in a blog post: "General relativity requires simulating the spacetime itself. That is, taking your simulation space, discretizing it to a hi-res 3-D grid and checking the effect that each and every point in that grid has on all neighboring points at every timestep. Instead of simulating N number of bodies, you are simulating a huge number of points. You start with some initial data of the shape of your spacetime and then see how it evolves according to the Einstein equations, which are 10 highly non-linear partial differential equations."
Why do I say this is not very satisfying or helpful?
Because it's quite easy to nearly completely account for General Relativity (GR) effects in the orbits of solar system bodies just by adding some terms to the ODEs that are integrated to propagate orbits.
Answers to the question How to calculate the planets and moons beyond Newtons's gravitational force? (including mine) show how this is done all the time. I reproduced what's normally done in Python rather than in a compiled language for transparancy, it would be much faster if included in a reasonable simulator implementation.
As explained here, to the Newtonian acceleration given by:
$$\mathbf{a_{Newton}} = -GM \frac{\mathbf{r}}{|r|^3},$$
You can just add this expression:
$$\mathbf{a_{GR}} = GM \frac{1}{c^2 |r|^3}\left(4 GM \frac{\mathbf{r}}{|r|} - (\mathbf{v} \cdot \mathbf{v}) \mathbf{r} + 4 (\mathbf{r} \cdot \mathbf{v}) \mathbf{v} \right),$$
To get close, you certainly don't need to evaluate these GR terms for every pair of objects. Interaction between the Moon and Ceres for example could still be Newtonian.
All approximations are approximate. The FAQ's suggestion that they don't do it because you need a supercomputer to be accurate is slightly misleading. Using this straightforward approximation is much better than not using it in this context.
Yes for Gravitational Waves, collapses, and other effects "requires simulating the spacetime itself. That is, taking your simulation space, discretizing it to a hi-res 3-D grid and checking the effect that each and every point in that grid has on all neighboring points at every timestep" but just adding the major GR effects to the propagation of solar system bodies certainly doesn't need anything more than whatever computer you are already running your copy of Universe Sandbox 2 on already.
EDIT:
The FAQ answer continues:
We are, however, interested in adding in a few features which would address some effects of relativity. One example is setting gravity to travel at the speed of light, instead of instantaneously taking effect as it currently does. You can read more about these in Jenn's blog post: Gravitational Waves & Universe Sandbox².
I had thought so as well, but according to @DavidHammen's thoughtful answer this is probably not so. I haven't tested this yet, but since this comes from a reliable source, I tend to believe it. Retarding gravity alone, without these other terms, seems to be the wrong thing to do in this case.
