Interesting.
My workflow is a bit complicated, so I will upload a video of a screen capture.
I work in stereo 3D, so I take L+R videos, and lay them one on top of another.
I reduce the opacity of top layer (L) by half to be able to see both and line them up.
I can then leave the top layer completely transparent, leaving only bottom layer (R) visible.
Once this has succeeded, I turn the two of them into a compound clip.
since playback is compressed L+R images, side by side into a single 1920x1080 frame, I distort the compound clip so it is squeezed into half frame.
You will see in the video that I jump to the project or timeline with this side by side distortion so you don't have to watch that process.
I then copy the compound clip, double click on the copy to reveal its component clips, and increase opacity of top layer (L).
I will then place this over the (R) clip, and also slide it to the left in the video frame.
Unfortunately, the change in top layer opacity in the copied compound clip has also applied to the original compound clip, so there is no difference between L and R frames.... both are showing the top, or L layer. (the lower layer of the lower compound clip which is camera R, has a noticeable colour and exposure difference and would be quite obviously different if visible beside the L compound clip)
am I approaching changing the opacity incorrectly? The reason for using compound clips is to be able to conserve registration for L + R images, when separated and put on each side of the frame...
here is a link to review process
https://youtu.be/sVlQBDbTzfE