The actual physics processing is done in _integrate_forces, but I set the flags to turn thrust on/off outside of that. I think this is where the discrepancy in the time-step is causing me problems.
I switched from 'add_constant_central_force()' to 'apply_central_impulse()' and it appears to have fixed it, but now I need to figure out how to integrate the "burn duration" slider back in. (it was just an await() to reset the constant force, but I didn't want to move that into _integrate_forces())
Looks like I have some refactoring to do now.
Thanks for your help!