Except in my experience most attempts at microservices scale extremely poorly because they have nowhere near enough precision in their APIs to get just the information you need and so most applications end up fetching way, way more data than they need and whittling it down manually - and often having to stitch together the results from multiple calls.

That of course leads you down the path of creating an "oddly specific" API that caters exactly to the needs of the clients, down to the point that small changes in the client's needs require changes in the upstream service(s) - so your "separation of concerns" has become a joke.