One video for alpha and one for RGB. Combine them in a shader.
If you want you can even have different alpha for each RGB element that way.
Should be simple and timing each video should not be a problem I think.
I wouldnt agree, you might have some problems of frame inaccurancies by getting two videos out of a hard disk. Why not merge both into one double sized video and then compute it by a shader? In the second thread link is a shader I’ve done for this purpose.