Seems like lately, from a user perspective, the current emphasis is mostly on doing the same things faster. So an indie browser developer probably won't get JS JIT down as well as a megacorp, but is that it? It seems to me like if you are standards compliant as of 5 years ago there shouldn't be a lot to do today.
Web browsers really are becoming similar to operating systems in scope.
Web push notifications are usually not an essential part of web app experiences and I agree they have done more harm than good so far. Some of these features enable types of apps that were just not possible on the web before. Things like 3D games where WebGL/WebGPU would be essential, text editors where they don't function without access to the file system, or a browser based video chat app that uses WebRTC.
As these features become available in most browsers their adoption will increase and any browser that doesn't support them will not be able to access web apps or pages that depend on them.
If somebody does depend on something obscure like javascript APIs for bluetooth [like someone in this thread cited], are they doing that as a desktop app with chromium embedded or are they a more traditional webpage? Does it completely fall over and explode if you don't have bluetooth support in the browser? That would be pretty unreasonable for most developers.
What I'm saying is there is a difference between the bleeding edge stuff or things that push boundaries and common practice. And if you are coding against that other stuff, you also need a fallback for if it doesn't work.
You basically need to handle an operating system where the VM is built on top of your browser, with all the security responsibilities and changing demands of developers working on your platform therein.