Connectivity is not always available, and errors retrieving it may mean you have to re-enter everything. Being able to "navigate" the site without the dinosaur is a tremendous boon for end users. You wouldn't be able to send anything without a connection, but you could buffer it for an opportune moment.
Sites like HN and Stackoverflow have a vastly different audience than most bread-and-butter sites, and tackle different kind of issues.
Most bread-and-butter sites are just fine as server-side rendered page. Large client-side frameworks reimplementing the browser and server functionality should be rare, when the situation actually calls for it.
If you really need a offline site then a simple service worker is all you need to cache the pages, not a big JS app.
Sites like HN and Stackoverflow have a vastly different audience than most bread-and-butter sites, and tackle different kind of issues.