https://arstechnica.com/cars/2023/11/rivian-apologizes-to-customers-after-infotainment-bricking-ota-update/
Some affected Rivian EVs may require physical servicing to fix the problem.
Interns do but should not get the level of write access that makes a durable change impacting all customers. Deadlock a server or even wipe SQL tables, this is an outage. Break a customer's configuration, send the wrong client's paperwork, again small scale problem you can deal with. Interns don't change company policy.
I think it's a more foundational architecture question: why do you push builds to all customers at once without gating it by SOMETHING that positively confirms the exact OTA update package has been validated? The absolute simplest thing I can think of is pushing to 1 random car and waiting for the post-install self tests to pass before pushing to everyone else. Maybe there's actually no release automation?? But then you make it safe a different way. It's just defensive coding practice, I'm not even a CS degree but learned on the job something always breaks so you generally account for the expectation that everything will fail by making a fail-safe just so the failure is not spectacular. Nothing fancy, just enough mitigation to keep the fuck up from eating into your weekend if it happens on a Friday.