Can you elaborate on the scaling issues? Being able to query into JSON fields (including functional indexes) is a great helper for denormalization which is very good for actually increasing scalability.
For me, the native JSON support is a very handy tool to have in the toolbelt for parts of our application that have a very loose schema.
Well your database is a black box technically speaking. It's very hard to scale it horizontally and it is very expensive to scale it vertically as time goes on.
Logic suggests that you should keep as much processing functionality outside something which can't be scaled cheaply or easily and push it to cheaper front end servers.
On this basis, anything which implies more work than collecting and shifting the data over the wire shouldn't really be in the database. Parsing / processing JSON is one of those things that's going to eat CPU/memory.
Fundamentally there's nothing wrong with storing JSON inside the database and processing it externally, but processing it inside the database is a big risk.
I've seen the same thing over the years with XML in the database and more recently people adding CLR code to SQL Server stored procedures.
For me, the native JSON support is a very handy tool to have in the toolbelt for parts of our application that have a very loose schema.