Interesting that this whole thread makes no distinction between read and write access, as those are dramatically different use cases. Read access is by far the more necessary, and is usually solved relative easily by saving snapshots to a data warehouse. This is no panacea as data can still easily be misinterpreted or replicated and used out of context in violation of expected production data lifecycle by the team that owns the source-of-truth, but in many cases it's trivial and minimizes coordination overhead.

Write access on the other hand is a different story. IMO restricting write access of a database to a single application is table stakes for scaling complex applications. Sure there are other approaches such as writing all your logic in the database via constraints and stored procedures and making the DBA a god-like figure, but these approaches have fallen out of favor as they've proven less scalable compared to wrapping a DB with a service that has exclusive write access. The latter arrangement allows many constraints to be enforced in a horizontally scalable and more legible layer, while still leveraging the DB to prevent races with a better menu of tradeoffs.

Of course this requires thoughtful service and interface design by competent technical domain experts, which is easier said than done, but the alternative allows the overall system cohesion to degrade to where no one understands the system well enough to make any changes without risking major incidents. At that point, the agency of system builders and maintainers is replaced by care and feeding of the unknowable system to not upset the status quo, accompanied with increasingly byzantine hacks and workarounds to enable any business changes.