You're considering network transmission time, by which cached JS is indeed free, but ignoring parse/JIT time and CPU usage as factors. Processors aren't getting any faster, and all that client-side rendering code has a very real cost in terms of compute-bound work for the user's hardware. Processors aren't getting any faster.
The usual answer I've heard here is to use an SPA, but those have problems of their own. They chew up huge amounts of memory if you use tabs, their UX is worse than if you'd just built normal webpages, you have to deal with making it indexable, and so on.
Honestly, I'd be ecstatic if most pages loaded in two or three network round-trips. A cold request takes maybe 100ms when I'm on a cell connection, but JS-heavy pages I've seen tend to have load times measured in seconds.
The usual answer I've heard here is to use an SPA, but those have problems of their own. They chew up huge amounts of memory if you use tabs, their UX is worse than if you'd just built normal webpages, you have to deal with making it indexable, and so on.
Honestly, I'd be ecstatic if most pages loaded in two or three network round-trips. A cold request takes maybe 100ms when I'm on a cell connection, but JS-heavy pages I've seen tend to have load times measured in seconds.