I am very happy you are willing to discuss my idea.
I will try to develop a little bit more, so you can get the theoretical framework I am thinking of.
You may have heard of the work of the art historian Aby Warburg, founder of an important library, The Warburg Institue, now based in London UCL. Back in the twenties, when photography was starting to become usable for "avant-garde" art historians - which followed the example of Heinrich Wölfflin - Warburg envisioned a "picture atlas" that had the scope of retelling the gestural history of the Occidental civilization.
If we keep aside his ambitious intentions, what still remains fresh and usable is his comparative methodology, which consisted in juxtaposing images of different size, media, chronology and provenance, connected by the idea of sharing the same gestures. You can take a look here and eventually read more about this enterprise: https://warburg.library.cornell.edu/panel/47). You can find bibliography also, but if you want to know more about the theoretical foundations of the method and its current use, I can write one about that.
My knowledge of this method made me aware that when we are (at least art historians, photographers, I guess also designers... etc.) dealing with image visualization on the pc, we are constantly obliged to visualize the information in a rigid structure, that is that of the list or the grid. This is what you get in any photo management software that I know, here comprised Lightroom.
So I believe that the possibility of manually moving the images is crucial because it would let people who work comparing large amounts of images to break the list/grid view with a single click (so without renaming the original files, or changing the rules that order the original list). Ideally, this interface would look similar to a normal desk where you can move around postcards (like those you buy in the museum). That's why I was thinking about the idea of the CAD-like, neutral plane, opposed to the grid/list view.
The overlapping feature would be fantastic, but it could be put as a second moment in the eventual development of this feature.
Ideally, a query into the database like: show all images tagged as "Florence" + "Michelangelo" + "Statue" would open all the records that verify this (maybe a trick in order to not overload the computing process would be adding an intermediate window where one could use checkboxes to decide what images are shown [queries with few tags could retrieve, say hundreds of images...]).
Then the UI would sort (as "free" columns or rows where the elements are movable via simple drag and drop?) the images in the CAD-plane following a criteria choosen by the user: the most common and obvious would be a chronological order, but I'd bet there would be others like author, color, original dimensions, media (show all the miniatures THEN all the sculptures that are tagged as "David"...), and most importatly: place of origin, a kind of geographical tag.
I elaborate this while working on my Master thesis where I was dealing with around 400 images that showed the same iconographical motif (in my case it was a biblical passage) from the X to the XIX century. I found it was very difficult to do extensive comparative research just using my memory and sorting the files in folders that I ordered with chronology. This seemed the most obvious mean but I could as well be curious about the chronology of that iconography in a certain region, or the same chronology but in a certain media. My questions were: where this motif has been originated? Did a style moved between - say - France and Germany via miniatures or via pictured glass (vitrails)? Was that tapestry effective into fostering the creation of prints that take the basic design from the brodery? I see these as kind of "meta-chronological" questions, that could be answered or at least evaluated once one could have a free hand on massive amount of digital image data.
I hope this description is more detailed, and I am more than willing to discuss with any developer also in private in order to develop the idea.