>instances of catastrophic failure are known by pretty much everyone.

I don't think anyone says we've seen the worst possible failures of nuclear plants.

As an extreme example, what if some crazy terrorist deliberately blows one or several up attempting to maximize damage? How bad would that be? How frequently does this occur?

I guess my point is that if there's real risks of that magnitude, they are just as invisible as coal radiation.

I can't envision a remotely plausible scenario that could even come close to Chernobyl in terms of impact. Chernobyl had no containment structure at all and was running at full power (infact, beyond full power once the steam inside the reactor increased the rate of fission) at the time of the incident. The steam explosion was so powerful that it launched the 4 million pound reactor lid through the roof scattering bits of fuel all over the countryside and then letting the radioactive graphite from the reactor burn and send highly radioactive smoke up into the jetstream.

Chernobyl consisted of an unsafe reactor design with a positive void coefficient and a control rod design that locally increased power output as it was being inserted combined with terrible management and operators who did not understand why the reactors power output was low and a complete lack of a containment structure to top it all off.

I genuinely can't envision a scenario like that happening with other non-RBMK reactors. Even if the operators were intentionally trying to just do the most devastation physically possible, it's not like they're going to be able to do anywhere near the damage that Chernobyl did. Trying to compromise the containment structure is not a reasonable goal, they are just far too massive and at least in the U.S. they are designed to take a direct hit by a fully loaded passenger jet without rupturing. Let alone rupturing the massive thick hunk of steel that is the reactor vessel inside of it.

There's still a couple of retrofitted RBMK style reactors operating in Russia but supposedly those don't have the same void coefficient problem, have better management and operational practices, and have better safety systems than the original design. Even if you count those against the safety of new nuclear power plants we're still not as bad off as Chernobyl was.

Thanks for your detailed response.

I hope you're right about the inherent safety of modern reactors, even in the face of malevolence. I don't know enough to contradict you.

In 1605 Guy Fawkes put enough gunpowder under the UK parliament to go through 7-foot concrete walls.

But if it's impossible to blow up a containment vessel, that's fantastic.

All I'd say is that I stand by the point that history to date is only weak evidence for future safety, and that as a general point, it's very hard to build something that survives dedicated adversaries.

Nuclear power plants are designed with scenarios like terrorist attacks in mind. The problems we've seen have all stemmed from systematic failures that were complex and hard to predict.

I feel folks aren't really taking the effort to understand my point here.

>Nuclear power plants are designed with scenarios like terrorist attacks in mind

Look, the world trade center was 'designed with scenarios like plane crashes in mind'. The towers still fell :( No one really expected bad guys to use planes as weapons, and things to get so hot.

Fukushima was designed with tsunamis in mind. No one expected a tsunami so far outside the design envelope.

That's fine, you can't economically design for every eventuality, so you define an envelope.

If you design a system to be robust against the 1 in 50 year Tsunami/attack/whatever, then after 50 years, all events are likely going to be in your margin of safety. If you extrapolate purely from that observed data, you are going to get a real shock 70 years in when a 1-in-100-years storm/attack/whatever occurs.

It's great that things are designed for robustness. But it then takes a long time for the true costs to be naively estimated just by looking at historical data.

Analysis based on 'the problems we have had' is fundamentally rearward looking. There's a value to it, but it's not the complete story. I argue its particularly dangerous with a technology that has rare but catastrophic failure modes.

In particular, it is flawed to look at a limited history of a tech with rare but catastrophic failures/costs, and empirically compare against a tech with more normally distributed failures/costs.

The companies that ran Fukushima failed to implement tsunami-proofing measures that were standard internationally, and studies had shown were necessary at the site [1]. So back to my point about complex failures: an engineering project's success is a function of both the physical assets and the organisations set to run them. Chernobyl is a prime example of where this intersection resulted in catastrophy.

I do understand your point, by the way. I completely agree that we can't account for every possible scenario using history alone. But there's a balance between over-engineering for fantastical events that are difficult to predict and keeping things affordable. Equally, at some point you have to weigh up the very present and deadly effects of coal power plants against the low probability, high impact of a nuclear power plant going boom.

[1] https://carnegieendowment.org/2012/03/06/why-fukushima-was-p...