But it's a matter of probabilities. It's not like a train will derail if it's over some length, and won't derail if it's shorter. A 10-car train can derail, it's just less likely at least for causes that are related to train length.
There is no regulation that will prevent all derailments with 100% effectiveness.
The normal way engineering is done is to create a safety margin so that the probability of a serious problem under normal operation is vanishingly small. Airliners do not fall out of the sky on a regular basis in the United States for that reason. Serious problems have become rare through serious application of serious standards.
Engineering rail systems so that derailments do not happen at all is a reasonable shorthand for almost never. It is engineering malpractice to run things so close to the edge that fatalities or serious property damage are to be expected next month instead of sometime in the next few decades or so.
That's true, but there are also engineering principles at play unrelated to probability. If I squeeze a piece of hard spaghetti on it's ends, would it be easier to break a 20cm piece of spaghetti or a 1cm piece of spaghetti? 20cm, obviously, and that same principle applies with trains. Longer trains are harder to stop, harder to control, less robust to disturbances. Likewise, if I squeeze a piece of soft spaghetti on its ends, it's harder to predict how it will bend if it's really long than if it's really short. You won't see a 1cm piece of spaghetti contort itself into loops, but a 20cm piece certainly will.
The spaghetti metaphor tells part of the story, though I’ve never seen spaghetti somewhat coupled (constrained) to some kind of path (analogous to a track).
> Longer trains are harder to stop, harder to control, less robust to disturbances.
All other things equal, this seems plausible, but I’m no expert on trains. So if I studied it, I would likely agree in most cases.
However, I do wonder if longer trains, being more massive, do offer some benefits in some cases. Perhaps a disturbance affecting only one car would have less effect because the other cars have more inertia and resist change.
This of course relates the question of the design criteria. Clearly trains should be designed with a cost benefit analysis such that more common disturbances can be handled. What kinds of disturbances can and should be planned for?
There is no regulation that will prevent all derailments with 100% effectiveness.