Regarding point 2: what if Tesla pushed silent OTA updates for A/B test that accidentally disabled ABS or traction control, rendering the car unusable? This is

Regarding point 2: what if Tesla pushed silent OTA updates for A/B test that accidentally disabled ABS or traction control, rendering the car unusable? This is what I cannot grasp about modern software project management/ops. <s> The only relevant measure is time to ship. If the unit tests pass it is good to go. </s> So many times have I seen webapps so broken that core functions do not even load or by no means would pass simple "click here and there" usability test. In house QA sometimes seems to be abandoned. Automated testing is a good thing, but in my book shipping comes with implicit assertion that changes perform the task at least on a basic level. Field A/B testing of experimental features is a good thing, which I am ok with even if I happen to get worse alternative. Shipping outright broken features if I am not on beta/testing/nightly release channel is unacceptable.