Apropos of nonlinear simulation, this is interesting:
The simplified case (point bodies in an ideal Newtonian universe) has confounded mathematicians for centuries now. I was pleased to note the the wiki article mentions the possibility of collisions, which make things more interesting. They don't mention relativity, gravitional waves, radiation pressure, or tidal effects.
If you include that stuff, even the two-body problem gets nasty.
LT Spice can have radically different runtimes and solutions if you make tiny changes to circuit values or time steps or initial conditions.