That's a pretty harrowing view. I see the movement of logic from the server to the client as liberating, personally, as it gives me more control over it.
Javascript is open source by its very nature. So if I wanted to tweak how Slack works, for instance, I can do so via extensions or writing code myself. In the cases of completely client-side apps, I can even back them up for my own use, such as I've done with a regex testing tool.
You've referred to it as a regression, but let me give you an example like Twitter. Would it make sense to have the server generate the HTML for each and every request, and then send that as a new page whenever you click a link? It's actually much faster to instead have the client send a small AJAX request (give me tweet #3242565), the server respond with that data, and then the client to update the page with it.
Twitter on its face is pretty simply. There's only a few template pages it needs to know ahead of time (timeline, individual tweet). So by moving to a data passing model you actually send far less data. The "javascript payload" being too large can be mitigated by rendering on the server for the first view, as is done in React.
In the case of Twitter, you're still dependent on the server for info either way. But this lets you conserve data, reduces server load, and yields a faster client interaction.
As I said, the web isn't static documents anymore. We've moved beyond that and it's actually pretty great. HN's overall position on this subject disappoints me for that reason.