The idea is to get whatever is causing the kernel panic worked out before you deploy the box outside a supervised lab. If you're still getting kernel panics, the code's not ready.