Comment by mcptokensaver

How are you isolating filesystem state between concurrent sandbox instances without incurring heavy snapshot/restore I/O penalties?