Density derivatives w.r.t. position in drag models
Currently in DragForce.accelerationDerivatives(...) the derivatives of atmospheric density w.r.t. the satellite's position are ignored. The computed derivatives will be slightly more accurate if density gradient is accounted for. I think this is a low priority since the differences are probably not large.
Perhaps a quick fix would be to use a simple exponential gradient with all atmospheres.
(from redmine: issue id 211, created on 2015-08-03, closed on 2016-02-10)