Thanks for the ideas, but now I'm not sure they would be very different from my original approach in terms of rendering performance. What seems to be happening in my game is that the so-called fillrate capacity
of my GPU is being reached. This is because I'm drawing a lot on screen already, and having to draw a white object on top of all the entire screen maxes out the device's rendering capacity for a frame, so it drops in framerate during fades.
if my understanding is correct, the new approach I use should strive to limit the number of pixels the GPU has to draw, so I don't use up my device's fillrate capacity. The approach Colludium suggested did this by 'flattening' the rest of the game content by taking a snapshot and then hiding the other layers. This could work, but I'm afraid of the huge image size and it sounds like a slight headache to implement.
So now I feel that a shader may be the best overall approach for fading. The other approach that comes to my mind now is to somehow "flatten" my game content to reduce overdraw, but I feel that the real culprit is a poor fade implementation. Clearly, games with reasonably complex scenes would have the same issue with using a sprite to fade. There must be a better way!
But yeah, thanks again for the help guys! I'm still open to new ideas, but I'm pretty sure fillrate is the issue here, so anything that uses sprites over the whole layer will likely not work.