It has to be both to work; the sandbox would fail in a day if there were no review/revocation system.
On the other hand, you can't be completely sure that sandboxes on mobile devices don't have actively exploited security issues as there are many ways to bypass app review from discovering the true functionality of an app.
Anything widely known gets fixed quickly. There are plenty of holes in browser sandboxing. The number approximately doubles as soon as you look at anything !Chrome, too.
To exploit those you need to get past app review in the first place, though, and the type of code that can do stuff like this for the most part sticks out like a sore thumb when subjected to static analysis.
Most of Apple's checks around this stuff are automated, I understand, and are applied to every submission instantly.
Basically, for certain classes of apps, macOS is now already taking the iOS "App Store or gtfo" model.