Its about efficiency.

Our input bandwidth is high through our vision and hearing in our current form. While this might change in the coming decades, right now our vision is the main information input channel. This channel is made up by a complex body infrastructure from which the first elements are our eyes. They are optical sensors, capable of image focusing, light intensity and color detection. That's it. That's what we can pass on to the brain, an image. All pattern recognition, information classification and matching is somewhat independant from vision.

Now, you propose that a '3D desktop' will be the future. Maybe in the distant future, when our input bandwidth in no longer limited by vision and we have a direct brain connection to the internet through some clever interfaces and protocols.

Right now, and probably for the next 10-20 years, we are limited to using our eyes and they can't input real 3D information. There are some perspective and tricks but you never have really a 3D vision. (Even '3D' movies and games are not really 3D. You see at most two images from different perspectives. 3D means full 3D view. See the whole object, front and back, top and bottom, all at once. This is 3D. A simultaneous double perspective vision is sometimes called 2.5D)

A '3D' digital desktop using a planar screen still is limited to show images, which is the physical objects our inputs (vision) can process, images. We have no way of inputting 3D information to our brains yet. And we have no device to output 3D information quickly either. Sculpting, machining, 3D printing, etc. can do it, but it's slow and very limited.

So, what is the point of a 3D digital desktop which is displayed always in 2D? That is inneficient and probably why the concepts were abandoned. We should optimize our desktops and output devices (screens) to fit our input channel (vision), which is limited to images.

EDIT: VR and AR are still outputting images, no real 3D. The bandwidth therefore can't increase much. The only way for a 3D input is through some new sensing interface, as I said, maybe a direct cortex-internet connection.

Amiga workbench was multitasking before windows!

Two eyes, two ears, two hands. Multi-channel music formats failed for a reason. The paradigm of windows layered on top of others is sound, but it looks like you're looking for a new 3D display device, where you can move your head and actually look behind things to see and bring others forward, rather than a new desktop formula. A rectangle with depth as opposed to a two dimensional one.

Anyway, I wish you luck and hope you come up with something amazing. :-)