Thank you. That was the first realistic assessment I've noticed so far in this discussion.
There is no programming language that is ideal for all contexts, nor any VM that supports all use cases well. There can't be, because there are some fundamentally contradictory goals that simply can't be fully reconciled. For example, you can't have a language that efficiently manipulates hardware for systems programming yet which also lets you run general applications downloaded from untrusted sources in a safe sandbox. The trick is to pick out the common elements and principles that are shared by some of the cases, and pool that work as much as possible so the maximum resources can go into making each tool better in its particular niche.
It's good that people who have grown up with "web apps" are starting to notice these issues that are old friends in the wider programming world, and that the Web world is starting to run into hard problems and realise that there aren't always easy solutions to them. This will inevitably lead to greater maturity in the technologies we use on the Web. But that doesn't mean the Web world somehow magically has answers that no-one else who's been working on this for the last 30/40/50 years has found.
Ironically, possibly the biggest lost opportunity in recent years is that Microsoft could have shifted the whole industry for the better just by having a decent, standardised install/upgrade process for native applications on Windows desktops. If they had produced something with the ease of use we now expect from mobile app stores, but the range of grown up applications we expect from a serious operating system and the freedom and support for developers that Microsoft used to champion for many years, then they might have disrupted the killer feature of Web apps -- which IMHO is the near zero friction ease of getting hold of them -- almost overnight. Throw in a decent set of tools for supporting client/server communications, and what advantages do Web apps really have left over native-but-distributed development?
Sadly, MS never really solved that problem, and so instead of today's new programmers growing up with a wide variety of native languages and run-time environments that built on decades of successful language and VM evolution, we get stuck with the joke that is Javascript, a bunch of server-side languages that all look much the same when you strip away trivial syntactic differences, and worst of all, insane amounts of development effort going into building several variations of ad-hoc, informally specified, bug-ridden, slow implementations of half of what a cutting edge VM should have been by now, all locked up in different browsers where nothing but Web apps can even use it. The rise of Web apps has probably regressed the software industry by at least a decade and counting. :-(