We've seen a couple positions put forward that I will paraphrase as "I personally see strong potential in Emacs for Android, although I don't/wouldn't really use it right away." As near as I can tell, these positions are tangential to Eli's point which I will summarize as "The Android port is great, and we should probably give it _some_ support, but maybe that level is lower than everything expressed and implied by incorporating it directly into the mainline emacs repository?"
I am personally sympathetic to both views, and (at the risk of forking/hijacking the thread a bit), I would even go so far as to say that Emacs as a project might benefit from spinning out at least the Android port (which is new and very maintained, but has a very high bus factor) and also the NS port (which is not new, shows some downsides of high bus-factor parts, and has at least one well-maintained alternative outside the mainline Emacs repo: the mac port).
As a technical matter, it seems like it's probably easier to maintain some abstraction barriers along OS and window-system code by virtue of having a single repo that supports 3-5 "variants", but in practice that _seems_ to have mostly resulted in treating a quite old "posix-y-unix with X11" as the baseline, and then adapting everything to that. This seems (again, I'm not an expert here) to have caused some continuing pain with respect to GTK and the pgtk port (particularly in the very high incidence of people using pgtk in wayland+X environments where it's not supported, kinda works, and breaks in not-so-rare corner cases).
Taking the NS port as an example, if the mainline were to drop support for the ns port, this would nudge some macOS users over to the mac port, strand some users of quite old macOS versions, and isolate the users of the GNUStep port. My belief is that the mac port is already quite popular, the people stuck on old mac OS X versions are already stuck with library and compiler version issues, and the GNUStep port has very little actual usage (I wouldn't be surprised to learn that "testing GNUStep support" is the single largest user-base of GNUStep" at this point.)
Before discussing whether these tradeoffs are worth it or not, is anything I've said above obviously wrong, under-informed, or out of date?
~Chad