A timestep of 1fs may still be too large for rigid bodies, especially with hydrogen atoms (see this post). You can try restarting the simulation from a state just before the crash, run with a much smaller timestep, and see if that helps.
A timestep of 1fs may still be too large for rigid bodies, especially with hydrogen atoms (see this post). You can try restarting the simulation from a state just before the crash, run with a much smaller timestep, and see if that helps.