It seems you missed the whole characterization of nothing is 100% safe. At some point there has to be an acceptable loss. Or rather, we have to try to do the least damage. But yes, people will die. But because we have electricity, more people will live.
I'd been unaware of Banqiao until a few years ago. Its scale overwhelms me. And yet the chain of institutional, engineering, and circumstancial events triggering it offer a huge set of cautionary lessons to nuclear advocates. I recommend reading the history closely.
Today, and for another 300 years, Fukushima and Chernobyl have effectively no inhabitants.
Zhumadian city, inundated by the Banqiao disaster, is home to over 7 million souls.
Once dam breaks cease being raging floods, life resumes, for those not fully extinguished, in a matter of weeks or years. Not centuries.
And the factors in assessing, avoiding, mitigating, alerting, and responding to risks are all remarkably similar to those of nuclear installations, save the very long-tail disasters.
The US has seen few major dam failures, though several have ocurred. Johnstown (1889) saw by far the most deaths, 2,200 (it spurred creation of the Red Cross and massive reforms to liability law), but see also the St. Francis (431 souls) and Teton (11) failures, and near misses or ongoing risks at Oroville, Isabella, Glen Canyon, among others.
Elsewhere, there are the cases of Vajont (2000+), Machchu (5500+), and others.
Again, the failures largely accrued from institutional hubris, engineering insufficience, lack of domain knowledge (often deliberate ignorance or denial), poor overall management, lack of disaster preparation, drilling, or readiness, communications breakdown (see Banqiao's comms loss), and inadequate resonse in light of imminent or present threat.
None of these are domain-specific to hydraulic civil engineering or absent from nuclear engineering projects.
