Is this guaranteed to work? The Chernobyl disaster was caused by grossly-improper maintenance, so… :shrug:. No matter how many physical safety mechanisms you have, a determined mechanic could remove them – but at that point, it becomes deliberate sabotage, which is rare, detectable, and not unique to nuclear power plants. (You could probably kill more people by destroying the right dam in the middle of the night.)
At scale, in the long run, you will get grossly improper maintenance. I feel comfortable asserting that without feeling like a total cynic.
It's not just nuclear either, this is also a problem with dams.
And that's before you take into account second order risks like people dying because they can't afford to heat properly during winter.
The damage from Fukushima has been so small that it wouldn't make it past regional headlines if it didn't involve the scary word "nuclear".
One solution is to house government officials and their families close-by so they're incentivized to never cheap out.
How many times did this happen in 80 years? according to [1] not many times.
[1] https://en.wikipedia.org/wiki/Lists_of_nuclear_disasters_and...
And this is including very early technology and government (e.g. the Russian one) that I wouldn't trust with anything.
The consequences of a nuclear incident are very high, and can be more or less permanent for an area. It's a lot to ask for a technology to be absolutely resilient to mismanagement or even sabotage.
Now apply this logic to many other power sources: https://en.wikipedia.org/wiki/1975_Banqiao_Dam_failure
It's a lot like trusting your private data to a company. Sure, Google in 2007 is pretty great, but maybe you have some doubts about their integrity in 2025. Too late, they have what they have, forever.