It still makes zero sense to take the XDG/dbus/whatever stack and make it run on a phone, suboptimally, when AOSP is right there and has already solved all the thousands of integration issues you'll run into --- plus, it's already free software.
NIH is the only rationale for the "Linux" phone thing and it's why it will be forever fringe. People working on "Linux" phones as anything more than a diversion (why not play Factorio instead?) are wasting their time.
> it's already free software
I'd just like to interject here for a moment. The word Free Software has a specific meaning that AOSP does not meet. The only component of AOSF that is Free Software is the Kernel, due to GPL, and aside from low-level Android-specific modules such as binder there's no secret sauce in Android kernels; even the vendor modifications are mostly gutted out in favour of Project Treble and GKI. Everything else is only Open Source and not Free Software, and even then developed privately and only published upon release. Because nobody releases a pure AOSP phone (Google Play Services alone changes the OS behaviour dramatically, punching through all the usual app sandboxes) and the source code for the modification, it's effectively proprietary with open source components.
I agree with your overall point but the following comment is unnecessary:
> People working on "Linux" phones as anything more than a diversion (why not play Factorio instead?) are wasting their time.
People are free to spent their free time however they want. Some people view building things, whether it’s furniture or software, more enjoyable than playing computer games or watching TV.
It makes sense because it gives you complete control over your device, to a level even AOSP can't touch.
One could ask why AOSP was created when there was an existing userland on linux that could be used (and was, by Nokia).
More seriously, I think the reason people want to do this is threefold: 1. Android vendors almost universally seem to make it hard to run stock AOSP (and do the Windows bloatware thing that Windows vendors were known for), so a "linux phone" lets people run what they want and remove what they do not 2. AOSP, while open source, is not developed in any way like a community open source project, so their ability to change anything, especially anything Google does not want to change, is limited and means constant rebasing 3. AOSP doesn't really solve the "run a modern/non-buggy kernel" issue on existing vendor hardware (as far as I know), so if you're going to spend time on getting the kernel to work, you probably want to have a userland that is amenable for getting the kernel working, so AOSP isn't helpful there, and by the time you've done all this, you can probably just run the rest of the standard setup with a distro and tooling you are already familiar with
I think the interesting thing would be if the modern kernel work from (3) could be used by an AOSP build and get the best of both worlds, or if by the time you do all this AOSP is too resource intensive to run on the device, and so running the alternative is the only option.