> The (legitimate) concern is that even low frequency accidents are frequent enough on the timescale of human civlization that they could present semi-permanent damage to the biosphere.

Any such concern, if legitimate, should at least be fleshed out with some basic assumptions and numbers. What is the (rough, estimate) probability distribution of the number of deaths that this will cause? Then we can compare it with the danger of fossil fuels. Otherwise we're left with hand-wavy expressions like "accidents" and "frequent enough". What are we to do with such statements?