[kragen]

when they're talking about probabilities like the one in 600 thousand of dying on this trip, they are the same

if you have a model where a probability goes outside the (0, 1) range for some input parameter value, that obviously isn't a valid answer, but it may just mean that the parameter value is well outside the model's range of validity. you are surely right that if we extrapolate this probability model of doubling the risk of death per 16 kph speed increase beyond about 400 kph we get invalid probabilities, but i think that probably the model breaks down well before that point