We tried building one last year. The tools, standards, and documentation are a mess making this unnecessarily hard to test and develop. The experience of triggering the install button is downright inconsistent and surrounded by layers of indirection, validators that have to line up and a complete lack of any meaningful feedback when it doesn't work as you expect. Cross browser support is basically missing. Firefox does nothing productive with PWAs as far as I know and seems to be in no hurry to fix that. I'm not sure about safari/ios since we did not bother testing with that much. In any case, it's indeed extremely rare to encounter PWA buttons in the wild on Android and it seems like mainstream web sites are completely ignoring this feature other than maybe as a gimmick. IMHO major work to address all of the above is needed to change that.
Major work as in making this a proper cross browser standard that results in people actually being able to use it instead of just talking about maybe using it some day when it stops sucking this much. The sad state of PWA is that if Chrome were to drop support for it tomorrow, very few people would even notice this and very few applications would stop working.
My impression is that devloping and testing this stuff is unnecessarily hard. The UI experience around this is flaky/inconsistent even when it works and the difference between this being broken/misconfigured or a user error is basically very hard to pin down. IMHO this is completely unusable in its current form and that's probably the main reason basically most people that ought to be getting some value out of this are either unaware it exists because they've actually never seen a real world example on a real world website adding any value whatsoever; or have made the conscious decision it's not worth their time (because of all the above). I'm in the latter category. I would use it but in its current form that's not something I can sell to a product manager.
