A bad hail storm comes close. Hail damage can total a car.

Cars are the cheap part of auto insurance claims.

Exactly this; damaging a building or causing the death of a person can be 10x+ more costly for the insurer.

Only when you are looking at one claim. If all the cars in a city get hail damage the total costs exceed the typical daily claim losses.

I think the point is that it’s much less than all the cars.

And a hailstorm that knocks out 10,000 cars is very rare. But hurricanes or fires that knock out billions in homes happen almost every year.

Euro importers love hail damaged Copart cars, very cheap to fix here.

Would auto insurers have enough insured cars within the area of a hailstorm to matter though?

This is why insurance companies pay cloud seeders to move thunderstorms and reduce the probability of massive hail claims.

This is true, we had a bad hail storm come through in 2010 that dimpled an appreciable fraction of the cars in the city like golf balls. Went right over a couple of luxury car yards. A bunch of people at my work moved our cars undercover 10 minutes before it hit, and felt kind of silly… for 10 minutes, until it hit.

Car insurance premiums jumped by quite a lot that day, as far as I can tell permanently.

I can easily imagine auto insurers facing exactly that kind of liability if a self-driving car release is bad enough.

> Auto insurers don't face a "catastrophic liability" bankrupting scenario like home insurers might in the case of a natural disaster or fire.

This changes with self-driving. Push a buggy update and potentially all the same model cars could crash on the same day.

This is not a threat model regular car insurers need to deal with since it'll never happen that all of their customers decide to drive drunk the same day, but that's effectively what a buggy software update would be like.

Far be it from me to tell automakers how to roll out software but I would expect them to have relatively slow and gradual rollouts, segmented by region and environment (e.g., Phoenix might be first while downtown London might be last).

That process itself could still break. (Unlikely though it may be)

Tesla certainly does it this way today. This is also the norm for IoT that I'm aware of. Nobody wants fleet-wide flag days anyway.

> Nobody wants fleet-wide flag days anyway.

Crowdstike raises their hand..

aionescu, CTIO of CrowdStrike, is here.

I think you’re right, but this thread did bring to mind the LA Northridge quake (1994):

https://scpr.brightspotcdn.com/dims4/default/a553905/2147483...