I think the separate QA before release step I bet was mostly killed by modern methods of software distribution. It's true that automated testing is great, but let's be real: automated tests have existed for many decades, and certainly overlapped with orgs having manual QA steps as part of release engineering.
If you're duplicating floppies, minting CDs, DVDs and so forth it's very expensive to make corrections. If you're pushing remote updates and digital downloads, or just running the software as a service and making it available over the network (e.g. a website) there's not the same expense associated with correction.
These methods of distribution also let careful organizations release to a small things out to a fraction of users first, which can minimize the costs and risks associated with a bad software push as well.
If you're duplicating floppies, minting CDs, DVDs and so forth it's very expensive to make corrections. If you're pushing remote updates and digital downloads, or just running the software as a service and making it available over the network (e.g. a website) there's not the same expense associated with correction.
These methods of distribution also let careful organizations release to a small things out to a fraction of users first, which can minimize the costs and risks associated with a bad software push as well.