Your points are quite valid.
I think we may have a small opportunity here, actually. There's been a fair amount of backlash in the Linux community over the directions both KDE and GNOME 3 are taking (over-complicated bloatware, inarticulate vision, and a seemingly arbitrary & piecemeal abandonment of their founding UI metaphors). I also get the impression there's a bit of a backlash brewing against the X Window system itself (which is one of the worst examples of combined specification and software bloat that you're likely to find anywhere).
If we can present what we want to do as a new GUI environment - which can run on any operating system, including OS/2 - that aims to recreate the Workplace Shell in its entirety (that means, including Presentation Manager) and is fully API compatible, then we might be able to drum up some enthusiasm.
By aiming at a self-sufficient system, we would implicitly need to start with the basic PM framework first – but if we can get enough OSS community developers onboard, we might even be able to have a PM development group and a SOM/WPS development group that work in tandem. (OK, maybe I'm dreaming in technicolour myself now!) But the thing about PM is that architecturally speaking, and in terms of the tools and expertise required, it's not really that complex AFAIK. That presents a lower barrier to entry.
I see nothing wrong with extending/improving the functionality of PM (for instance, adding some desperately-needed Unicode capable controls) as long as we are absolutely strict about the backwards-compatibility requirement. Indeed, IMHO we'd need to allow for some modernization in order to motivate people to get on board.