> destroying huge areas for good
Even Chernobyl isn't destroyed for good, though it's still an expensive mess to clean up. Fortunately, nobody builds reactors with such an idiotic design (a positive void coefficient[1]).
The actual site at Fukushima Daiichi is going to to be a long, expensive cleanup, but that's about the extent of the damage. Concerns about the surrounding area is just radiophobia and an extreme overreaction by the Japanese government. With how incredibly destructive the tsunami was, Fukushima Daini (which was also badly damage by the tsunami) showed us how incredibly safe nuclear power can be.
> can tell you that from looking at CS
More CS people need to study the lesson of incidents like therac-25[2]. It's very important to learn what fail safe means, which is very possible to implement, usually by avoiding as much of that "complex system". I trust the inherent and engineered safety systems of a nuclear reactor much more than I would trust any software project - including my own.
> A complete failure/breach of a big reactor is very costly.
So are fly ash slurry spills. You're holding coal to the same standards, right?
[1] http://www.hiroshimasyndrome.com/chernobyl.html
[2] http://sunnyday.mit.edu/papers/therac.pdf