I have also been working in this field for last few years. We do not have any analytic theory to solve "gravitational N-body probelm" so the use
of supercomputers seems to be ultimate choice.
There are many drawbacks of this approch. Computers work like "black box", we do not know what "actually happens" when a system evolve.
Everbody use his or her own "codes" and "machnines" for simulation and there is no standerization so there is little hope to compare results of one simulation to another.
Last but not lest in place of trying to understand physics, most of the time people just compete for bigger and bigger simulation and it cause wastage of governments money.