For early stage startups, it is invariably "none at all": You build it you run it. But that's doing two jobs, so you better be good at running it, and ideally you use dead simple technologies (none of this "k8s cluster-per-environment for a team of 2" business).
For very large companies, this may be public information. Do we know for sure Netflix has "Option A" (good team)? We expect Google to have Option A, but maybe with red tape galore (see yesterday's video).
But how about mid size companies? I have worked places with "Option B" infra teams. It was miserable, and I didn't last. Putting aside the question of how anyone puts up with that for long, how do we figure out who these companies are? What are the warning signs? Is it a situation worth fixing (maybe for a staff engineer, probably not for a mid level; as product engineers/outsiders, not infra engineers)?