Motivation behind not being able to get IO's content, and IO-wrapped interfacing with unwrapped (normal) code

Issue で言及されているドキュメントと IO 関連のエントリポイントから始め、特に unsafe_perform_io と、ラップされた Python コードと標準の Python コードの違いを確認してください。IO の内容を隠す理由を明確にし、IO でラップされた操作と通常の Python コードの間でサポートされる境界を文書化してください。

Hi! Question(s) on IO.

I am trying to write a little wrapper for some stateful operation (imagine I have to do IO from/to a file), and I found out that IO containers are designed to hide their content. It's impossible to get the value of, say, a successful read result (nothing like IOSuccess.unwrap()), unless one uses unsafe_perform_io.

What is the rationale behind this design? I am asking because I am not familiar with writing typed functional code, and I could not find a motivation in the documentation for why it shouldn't be possible to get the raw content of an IO container.

Second question related to interface between IO-wrapped code and normal "unsafe" code:

Let's say I am writing code A where want to mark and wrap all IO operations. Now, code A has to interface with standard python code B. What choices do I have to make them talk?

Do I have to unwrap all my operations with unsafe_perform_io, in case I want to do something simple like getting a string from my wrapped code? Do I have to rewrite code B by making it all safe and IO-wrapped, so that they can communicate with no "unsafe" bridge between them?

It seems that in such scenario I would end up with code pieces that are of "different colors" - using the usual async-vs-sync metaphore of incompatible red and blue code -, and I would be forced to make everything of the same color.

(Here "red code" = "python code written in railway style" and "blue code" = "standard procedural python code").