Why is is a binary choice which lens to use anyway?
They should be taking a photo from both lenses and computationally merging them. Sure - very close shots will mostly use the output of one lens, and distant shots will mostly use the other, but there should be no sudden switchover and there is no situation in which there is no value from extra data collected from the other.
I agree that the binary switch is very annoying - when zooming with my fingers I have to make sure I don't get stuck at 3.8x or something and need to cross over to 4.X to get decent quality. Ideally 3.8x would still use 4x camera to fill most of the frame and the main camera would fill just the edges.
The idea is quite simple, but it's rather difficult to implement in practice. The lenses have each different character, optical flaws, color rendering and naive merging would look pretty bad. But it does sound like something AI tech could handle.
They should be taking a photo from both lenses and computationally merging them. Sure - very close shots will mostly use the output of one lens, and distant shots will mostly use the other, but there should be no sudden switchover and there is no situation in which there is no value from extra data collected from the other.