I think the concept of creating a way for a portion of end users to develop human readable code that can be shared with other users is a great way to be able to create custom applications with a many magnitudes lower barrier to entry than learning to create an AUv3 app oneself.
I think DIY MIDI controller apps have a history on iOS but none of them have been translated to AUv3 format. The ability to tightly link via AUv3 host sync, save state settings, AUv3 automation, and running multiple instances are some of the primary advantages of an AUv3 app.
Currently, Audulus 4 is being developed for release in 2019. It will be a visual programming language for music including MIDI out functionality which will also be able to run patches in an AUv3 window.
The biggest disadvantage I see with Mozaic as proposed here is being limited to one GUI. Mozaic should certainly be easier to learn than Audulus 4 though I think in either case, the willingness of users to develop patches which end users find valuable will drive the degree to which these apps will be adopted. Audulus 4 will also have a much wider scope of functionality too so I see both of these apps having their own market rather than taking bytes out of the same Apple pie.
This seems like an excellent way to leverage the talents of the more skilled and committed professional music app developers (a rare and under appreciated resource) along with users invested in an app who will develop useful patches/code for the benefit of the app’s community of users. This in turn should reduce the need for professional developers to create as many apps and get some relief from feature creep as the community of users will be able to take some initiative upon themselves while adding value to the app.
Similar dynamics are in play when sound designers develop patches for synths although I believe the markets for apps like Mozaic and Audulus 4 may be smaller unless the user community embraces their efforts sufficiently.