Another thing, when I call print(GCMouse.mice()) inside the immersive space, it returns an empty array which is incorrect. I threw another print at the ToggleImmersiveSpace button before the call to openImmersiveSpace, and it shows an empty array when the button is pressed, and then at the second print statement (the same one from before that is still inside the RealityView body closure) it correctly shows that their is a mouse connected. Also the onDisappear is able to run correctly.

After looking at the backtraces, I see no indication of any thread being blocked. It is pretty similar to the case where no call to GCMouse/GCController is made and the onDisappear works as expected. As a note, this issues also occurs when the AVP is connected to a Mac Virtual Display, presumably since the trackpad can be used as a mouse input for the AVP.

I developed an app that would display the states of controller inputs as text inside a window view. But when I am looking at the window (even though there are no UI elements to interact with) then the states do not update. I noticed that if I direct my gaze away from the window, then the button values start to change and reflect their actual state.

Unfortunately, remapping may not be sufficient. There seems to be an issue with the joy cons and I have filed a bug in feedback assistant. I was hoping it was related to this question but maybe not. The issue is that none of the controller inputs work when the gaze is directed at any visionOS window. I had assumed it was related to the joy con being used to control said window but if there is no way to deactivate that, then it might be a bug.

Thank you for this response. This makes sense to me. However, you say that the update block on the RealityView is marked at @MainActor but unless I am missing something, this is not done explicitly?