Threw together a quick demo to show the variance possible in the shaky cam setup I described earlier. It sends the shame shake vectors to all cameras, the different results are due to different spring constants and different rigidbody linear dampening values.

Try it here