Apple fixed this and at least in the latest simulator you can have a task to dismissimersivespace and then openimmrsivespace. The fix appears to have been to add a small delay to the dismiss. Therefore if you want to immediately switch scenes it is more performant to have scenes in a master scene and load that master in RealityView make and then use update to switch what the user sees. Of course if you don't mind a kind of fade out and fade in effect you can dismiss and open.

And I think interesting immersive experiences could also be created for non-disabled users as well if this feature was available. Sometimes just a look could be sufficient.

In Unity for windows, for example, I am able to determine when the mouse is hovering on a 3D object and I can take actions on a timer. This has led to a significantly better experience for eye gaze users vs having to dwell click.

I create accessible apps and being able to initiate actions onContinuousHover in immersive spaces would improve the user experience for those who would otherwise have to use Dwell to tap. So RealityView entity recognition in onContinuousHover could be very powerful.

could you post your code with the manual delay that worked?