Note that by using all the extra nodes, you may end up with a solution that is less performant than straight physics implementation… 
This runs in separate thread on a fixed interval:
Still somewhat experimental so thread with care one way or another.