They both remain good ideas, but I think the Mylar bag should win out, as it would be able to fit more imactor scenarios than planting a rocket engine on the surface.

For the in situ rocket engine, calculations on when the burns should occur to achieve a benign trajectory are not that difficult to run.

For the Mylar bag, the thrust can still be applied, even as the rock -- or the conglomeration of frozen particles -- rolls under its weight. There will be a lot of momentum, but the structure can be devised to function as a catcher's mitt to allow the ball to continue its rotation, even after being caught and redirected.

Dan