> 1 defect in 10,000 is not arbitrarily 'good'

No, I'd argue it's pretty awful. The average person probably interacts with the police ~10 times during their life. If 1 in 10 000 cop interactions ended up with you getting killed, police brutality would be in the top 10 causes of death in the US (on par with suicide).

Temporal said 'criminal', not 'fatal', and gave the example of 'maimed or fatal'. There are also non-fatal, non-injurious ways to be criminal in policing.

Technically yes, but they also gave an example of a serious, non-fatal, non-injurious way of being criminal in policing ("terrorizing innocents at gunpoint") and classified that outside of the 1 in 10 000 frequency.

I didn't mean literal "criminal" as in criminal law vs. civil law, I was aiming for differentiating between things that seriously stress you up at worst, and things that leave you (or your family, or your dog) with damaged health or property. Bad choice of words on my part.

> 1 defect in 10,000 is not arbitrarily 'good'

Phew, agreed -- my jaw dropped at grandparent's 'minor'.

> we don't collect statistics nationally. So, how do we improve?

I think there's plenty of open-loop approaches that could be considered (training on de-escalation, refresher on citizen's rights, praise/laud the interactions that result in no arrest or arrest without aggression). Waiting for national consensus is unnecessary: cities, counties and towns can recognize the crisis and establish best practices that are effective.

National consensus is likely useful to allocate funds that may be necessary to make these improvements, though.