The fix that allows multiple force models to run with different settings on a single atmosphere model, seen in the "SchattenWeird" test case on
GMT-5323 was to assign an internal atmosphere model to the force model rather than to use the atmosphere model assigned to the central body. That change produced an undesired side effect in the MarsGRAM model, so a special case was added to DragForce to treat MarsGRAM2005 the old way (and retrieve its instance from Mars).
This approach will likely lead to the weirdness seen in the Earth models unless it is fixed. However, the model is not used extensively, so I marked the issue P2 – let's fix it when it is needed.