We tried going that route but it meant we didn't control enough of the stack to avoid issues related to updates etc.

Bugs would arise. Didn't work out.