This is the kind of feature I would be very nervous of implementing as a developer (showing some data in the context of another user)... to the point of pushing back on it. I imagine it cuts across whole swathes of code, requiring additional checks about the current user. It would be so easy to leave some of these checks out or mess them up. You would have to be very careful about validating your assumptions with thi…
I built an internal web application for a large and somewhat paranoid company that required three distinct layers of security. While I'm pretty sure I implemented them correctly, and could explain pretty well how they worked when pressed, I think it was just too much for most users to wrap their heads around the mental model.
In retrospect, I should have pushed back. If I couldn't convince them to with a simpler model, I should have at least advocated for adding the layers of security incrementally, both to let the users understand what the model is and to help the development team understand what the model should be.