Consider the barriers to browser performance:

Hardware acceleration - already there with WebGL and for video codecs The JS language - this can be swapped out for something more performant - various options exist but it will take time and a brave vendor to do so. Rendering speed of HTML etc - this can be improved, both by simplifying things like the box model and by improving the renderer, we've already made huge strides int this area for standard UIs.

These are solvable problems - eventually web properties could easily be compiled (in a sandboxed language or VM), and become native. I don't think there's necessarily a difference between a sandboxed process in a browser VM, and a sandboxed process on the desktop. Just because browsers are still developing and in the past have not been performant in comparison to native APIs, doesn't mean things always have to be that way. So while the present performance comparison of native to HTML means native wins (as long as you don't just do everything in webgl say), things don't have to stay that way.

The greatest strength of the web platform is that it is simple, open, and extensible. This also makes it a bit of a mess of competing standards and leads to a problem with legacy cruft (like the reliance on JS), but it has held up remarkably well, and has proven flexible, simple and incredibly popular. It's difficult to charge people money for content on the web, but not impossible and that's more to do with culture than technical challenges. It's also difficult to lock people in on the web - that's something people hate about native which isn't going away.

Eventually I expect the internet, and the web in particular, will reduce operating systems to a badly debugged set of device drivers (Andreessen was prophetic in that regard), and operating systems will be replaced as a concept and become simply the interface to hardware, which implements some standard APIs for the web and acts as a container for the VM which runs the code. We're not far from that situation already, and it is preferable for consumers and creators/developers. The only thing standing in its way are the vendors of large ecosystems (hardware and software) like Google, Apple, MS and Amazon, who have a vested interest in corralling and controlling their customers and keeping them inside an area where they set the rules and can take a cut on every transaction.

I'd argue the resurgence of native mobile platforms is the last hurrah of native, and simply another attempt to lock people into an ecosystem controlled by one company, it will fail as people start to realise their loyalty to say Apple is not rewarded. We'll probably see these companies grudgingly adopt the web, while still trying to corral their customers (it's what companies do when they get to a certain scale), but on the web that is much harder to do, because it's an open API, controlled by no-one.