I realize that with the recent developments here we can:
interpret at run time some C# parts of your application while compiling the rest ahead of time as usual.
This has a lot of benefits, and enables us to use packages like ML.NET to perform Machine Learning, and Roslyn that allows you to compile C# offline (without using a web service).
I know there are several limitations right now that are mentioned in the article, such as a really big package sizes. But is there any reason that an app built with features using the interpreter would be rejected during submission?