This has been attempted multiple times, and always fails because followoing FOSS to the letter doesn't play with how hardware industry works, and when people aren't willing to make concensions they cannot ever deliver a product the general public would replace their Android/iOS phones with.

GrapheneOS and SailfishOS focus on a narrow set of devices and they can keep up with hardware support. I agree that you have to make concessions in terms of allowing proprietary firmware blobs and opaque baseband hardware. You also have to choose your hardware wisely (e.g. GrapheneOS can/could piggyback on Google's driver work).

I was just saying that you can make the problem more narrow by not trying to support every device out there. Start small and pick your battles (which probably means using AOSP and using sandboxed AOSP).

I think the main issue of many previous attempts was what typically happens in the FLOSS community: there are N attempts rather than one coordinated attempt (Ubuntu Touch, Plasma Mobile, PostmarketOS, PureOS, etc.) and everybody is targeting different hardware. It's similar to how the Linux desktop got fragmented, though it's even more problematic for mobile, since the usage is probably 1/1000th of Linux desktop usage.