"Shouldn't" is what causes the question to fall apart, but swap that for "normally doesn't" and then it becomes accurate. People dying is usually the flag for when the safety system is invoked and then evaluates and responds. Pretty much all we're capable of is reactionary actions and even then it's pretty hit or miss.
To do what people expect, you'd need thousands of people in commercial buildings across the country sitting at their desks with three shifts of those folks working on every possible combination of roads and airspace and sidewalks and train tracks and subways, looking for all combinations to activity which could lead to something going wrong, and then proactively taking $tep$ to correct it.
That's just not what we do here. Pretty much every rule/law/guideline we have is from something happening or nearly happening and then someone trying to fix it, not so much from studying and hypothesizing to create scenarios and then protect against that. It's probably like a 1000:1 ratio or better (worse) of reactive to proactive.