Alright, it's your turn to be an engineer. Suppose you created a toaster, then found out that in service the heating elements burned out. What would you do?
Back to the drawing board. Hmm, what if we put in some kind of heat sensing device and automatically turned on a powerful fan when the heat got too high?
Well, fans are expensive. We could just put in a simple thermostat and set it to short out the toaster if the heat got too high, and depend on the house wiring's fuses or breakers to protect us.
If you chose either of the above, it's back to engineering school with you - both of those ideas really stink out loud. In fact, as millions of people worldwide can tell you, it's possible to design a toaster that will be almost idiotproof day in and day out for years without burning out or otherwise failing electrically, and the whole device can be had for under $10. And if you want to design something that will burn out at a specified current level or higher, those can be had even cheaper - they're called "fuses".
We can design this way because we can calculate the heat generation rates, the rate of heat transfer out of the heating elements, and the temperature at which the heating elements will fail. In essence, the heat generated will equal the heat discharged at some temperature T, and if T is less than the temperature that will significantly impair the performance and life of the heating elements you have a workable product.
It seems pretty obvious that you'd want to design a heating device such that it couldn't destroy itself by overheating if that was at all possible. You might think that that was especially important in something like, say, a nuclear reactor.
It is. But unfortunately there was a big push to get nuclear reactors working in the early days, and the first designs they got to work didn't meet that criterion - they were capable of getting hotter than what they could physically withstand. Lots of training, auxiliaries, testing and emergency planning are used to make up for this design characteristic. And really, these reactors were designed for submarines - all the cooling water you could ask for was right there outside the hull.
But then there was a big push for land-based reactors back in the 1950's, so submarine reactors were scaled up in size to make US commercial nuclear power reactors. The inherent overheating capability problem remained.
Now there's no arguing with experience. The engineered safety features and other measures taken by US commercial nuclear power plants have been very successful and safe by any reasonable standard (don't compare them to Russian designs like the VVERs or the RBMKs - the latter made Chernobyl famous and have far worse design issues than the ones discussed here).
But they make for a more expensive and ultimately less reliable plant than what could be realized with fuel that can't destroy itself or its containment by overheating. And these plant designs exist.
This is a technical reference on the plants.
Of course there are organizations that bitch about nuclear power plants for a living, such as NIRS. This is what they have to say about the design. This post is already long, so I'll Fisk the NIRS link later.