We have decided not to address this issue at this moment. The mass ratio is outside of what can be expected reliable simulation from. Also, from a theoretical perspective both behaviours (pre-change, and after change) are incorrect: before the change the large body would pierce through the small one and only accidentally drag it together at around exit from collision, after the change it won't pierce through but there might be arguably questionable reciprocal motion. It's a consequence of an otherwise good change that addressed CCD objects tunnelling in certain conditions. We might reconsider this resolution later.