I finished a rough proof-of-concept for abstract commands on top of command remapping. I used a macro to generate a mass rebind _expression_. I can still use C-n and C-p. I can rebind the abstract commands and all of the other commands follow. If I add bindings to #'abstract-next to C-; for example, all of the re-mapped commands are available on that sequence and shadow appropriately.
The tests also demonstrate the outputs, but are less realistic.
I am currently using Emacs with all of the mentioned keys rebound by the macro call. The first example of growing pains I found is that Ivy actually already uses a remap to next-line. Since it wasn't a built-in Emacs map, I didn't include it in my example macro. I would have to work on the macro and _expression_ generation to handle remap detection when performing a move, but this is more a matter of plumbing than complexity.
I was able to use my reporting functionality from the user-keys package to identify maps to start working on. The commented expressions in the macro caused errors (binding in the widget-global-map affects the global-map?) or strange behavior I don't want to invest time in before validating the use cases.
What I would recommend for Emacs is to implement a first-class named-key that doesn't have the implementation limitations of the remap only going through one layer. I would be able to focus on making user-keys into a better diagnosis and generation tool if I avoid going too deep on top of the remap-based implementation, which has the one-indirection depth limitation.
I'll publish the package pretty soon with some small coherence improvements. While looking at emulation maps, I realized some of them won't have symbols for example, so I need to figure out a way to represent the idea to the user. I've worked a bit on generating mass unbinds by refactoring the reporting logic and it generally looks like there won't be difficulty.