France is choke-full of nuclear power plants. Can you remember a major nuclear power incident in last, say, 30 years?

I think that reactor standardization really paid off there. They have few types, a wide operation experience, and apparently well thought-out procedures.

This, of course, is hard to achieve without a massive rollout planned ahead.

One thing I like to remember is that there are still 10 RMBK reactors in operation. Fundamentally the same design as the reactor 4 at Chernobyl. In fact, one thing a lot of people don't realise, the 3 other reactors at Chernobyl were restarted and ran until 2000 (actually, reactor 2 shut down earlier because of a non-nuclear accident with its generator) when the EU paid to have them shutdown.

The Chernobyl reactors were all perfectly safe. It was almost impossible for them to suffer a meltdown. That's the problem here: greed and incompetence and pride can't be fixed with better cooling systems. Today's perfect reactor is just as dangerous as the perfect reactor at Chernobyl once someone wants a bonus or needs to improve efficiency or doesn't want to admit he is confused and turn the thing off...

Perfectly safe is a long stretch. As originally installed they had numerous design faults, the control rods operated too slowly, the carbon tips of the control rods could locally increase reactivity when they were being inserted, the reactor was too large and consequently hard to control, they lacked the secondary containment structures which were common place in western reactors of the era.

> nuclear power is tightly coupled and complex

Sounds like you have read Charles Perrow's 1984 book Normal Accidents [1]. New fuel systems are actually key to safe design, IMHO. All the major nuclear power incidents, that I'm aware of, are due to one underlying problem: zirconium clad fuel rods produce large amounts of hydrogen during loss of coolant accidents (LOCA). Nuclear power plants weren't supposed to go BOOM but it turns out they do so readily with hydrogen explosions.

Even with zirconium clad fuel rods, the CANDU reactor design seems to have managed the complexity and tight-coupling issues surrounding nuclear power generation; the design, however, never managed to reduce the immense capital costs involved.

Price-per-Performance is all that matters and in an age of cheap natural gas. I'm not convinced that new nuclear reactor designs will find anything other than niche markets. There are many lessons to learn from Charles Perrow but complete distrust of engineering should not be one of them.

[1] https://en.wikipedia.org/wiki/Normal_Accidents

I feel like the real issue is something that plagues many complex systems: very smart people design something that fails to take into account the very fallible humans which will be responsible for actually running the systems. Scientists and engineers seem to overestimate the capabilities of humans to screw things up. Systems should be designed to be somewhat 'overdamped' so that lack of control results in the system shutting down safely.

My gut is that you can run a nuclear plant safely for 100 years, but as time goes on, you have fewer guarantees, and eventually something weird will happen to screw up your plans. I don’t think nuclear fission as we have it today makes sense except as a bridge to somewhere else.