The cost could be that much - but we've had 2 bad incidents, both considered "worst case", without anything near that situation. That is only 2 data points, an

The cost could be that much - but we've had 2 bad incidents, both considered "worst case", without anything near that situation. That is only 2 data points, and things could be worse - but that is the prior we have. How to weight a .00(0000)1% chance at tens of millions of affected, versus the guarantee of millions of life-years made worse by the actual coal plants? That is a philosophical plan, but the likelihood * badness is one approach. We also appear to have good control over the worst case situation with a plant - it looks to me like a bounded problem space with modern nukes.