Thanks Al. We find that the potential for harm comes from a multiplication of two factors: intensity * exposure. Our implementation meets R9m via minimizing the potential for harm by reducing the 2nd factor, exposure. We reduce exposure across both the space and time axes, by both (1) pointing the Limelight away from the eyes of field bystanders and (2) only lighting up the LEDs while targeting the vision target, achieved through the recommendations made by the helpful folks in this thread.