Someone attached an AirTag to my car while I was in a bar

The amount of effort they would put into stalking is mind blowing.

To me that sounds somewhat different from irrational rational fears and also as somewhat reasonable. Irrational rational fear attributes greater danger to an unlikely event due to ignorance of how unlikely that event is. While Nassim Taleb's thesis (AFAIK) acknowledges the events are extremely unlikely but posits a system is more robust if such events can be handled. That seems to make some sense to me, but I would be concerned if such counter measures become prohibitive. For example, I suspect we no longer need to restrict knives on airplanes. No plane is ever again going to be hijacked with just a knife. We should though continue screening for bombs and plastic explosives, an attacker would be more likely and be more interested to depressurize the cabin and try to crash the plane and kill those onboard via a bomb than to hijack it (where everyone now knows they should rush a hijacker no matter the cost).

A good analogy, a computer system that can handle errors that occur in 0.001% of cases is more robust than one that cannot. Particularly with large numbers thrown at it. The contra-point though is that real life is not large numbers, worrying more about a 0.001% probability compared to a 0.1% probability is generally misguided, particularly when a person believes that 0.001% probability is more like 10%.

Though, that is where conditional probability kicks in and here is why I agree regarding "damn lies and statistics". Hence, even if the odds are 1 vs 500 or 1 vs 5000, these are all negligibly frequent events for the average person. Though, if you are over 85, in a bathtub, perhaps on some new medication, recently recovered from a surgery, and there is nobody else at home, then I would suspect odds of dying in that bathtub are far greater than they would be otherwise.

As another example, it's kinda the same for Texas Hold'em, the odds of a 4 of a kind are generally low - but if the first 3 flop cards are all the same rank, the odds of someone having a 4 of a kind suddenly become reasonably quite good.

So, I think the interesting conversation is around conditional odds and when as humans we accurately recognize those conditions. For example, many cyclists ride to the extreme right of roads fearing getting hit from behind. The statistics for that do not support that behavior, it's really rare for a driver to hit a cyclist from behind ("eyes on the road"). It turns out the majority of those kinds of collisions occur at night and when the cyclist has no lights on their bicycle. Meanwhile that misconception of the danger puts the cyclist in a more dangerous location, the further to the edge of the road the less visible from turning traffic (the real danger to cyclists), greater danger from road debris, pot holes, bad pavement, and getting doored and pedestrians popping out from between cars and getting doored (tl;dr, it's much safer for a cyclist to be a couple feet into the lane than immediately adjacent the gutter). Though the irrational rational fear has that person riding next to the gutter, putting them in greater overall danger.

Yes, because unlike in Europe, people in USA have the right to shoot you dead if you're trespassing in their property.

A bathtub you have control over may not be especially risky. The stat itself probably includes unsupervised kids, elderly, etc. Take those out because they don't apply to me.

You don't have as much control over strangers. And any given homeless person may or may not be judged as especially risky - that's up for the person there to decide. There's also little downside to just avoiding them entirely.

To me, it seems like it can be completely rational.

Individual events don't happen with dice like you seem to be implying.