To stay obvious, what's fascinating is that those networks are small, its members the most intelligent people available and they meet each other regularly in person at conferences.
Not every community does it this way. For example, computational linguistics put most of their conference proceedings online for free: https://aclanthology.org/. Deep learning researchers just publish a lot of stuff to arxiv.
Academic publishers like Elsevier are predatory scammers.
Yep if something in CS and adjacent fields isn't open access (or there's a pre-print floating around somewhere) chances are it's a textbook, not worth reading, or is obscure/arcane and was written with a typewriter. Heck some of the best stuff is blogposts by people who don't happen to be in a publish or perish situation so why bother with journals. (Trouble with that, of course, is a lack of doi but what's archive.org for).
Meanwhile there's fields which can't even figure out TeX.
They may be intelligent in their fields but that doesn’t mean they think thing through in every aspect of their lives. The status quo is the easiest thing to deal with they can devote more time to their careers/research
Unless their field is in social engineering, then yeah why are they going along with it?
Because they need funding. Research projects take a lot of capital. And you'd need a lot of money to set up an independent journal, facilities, labs, staff, etc.
Like the other response to this said, it's a little more complicated than "the status quo is easier" or "intelligent doesn't mean smart." This is a deeply ingrained system that's existed for a long time, and if you don't operate within it, you don't get to work in academia. You won't get to conduct your research to begin with, much less will you get to the point of publishing it without cooperating with these institutions. There are also powerful regulatory bodies like the APA and AMA who control just about everything in their field. You pretty much have to work for a university, and US universities are of course greedy and corrupt in their own right.
It would be like unseating the DNC, ending the electoral college, and expanding the two party system in America, but on a smaller scale. Plenty of Americans know that these things need to happen, but it's not something where you can just wake up one day and make the decision to overthrow the system as long as you just try real hard.
Looks like there is no good answer if we view them as one entity which could simply make up it's mind. But it's a bunch of individuals, who probably disagree at least over details. Some probably have individual ambitions or pressures, some may struggle to pay their bills or satisfy their family or even themselves.
And for each individual on the fence, it's always an advantage to still publish to the network while hoping the rest of the group abstains and establishes a better platform in the meantime. Would you risk publishing your finally successful hard work to an immature platform, where it might not receive the attention it deserves?
And because they're smart, they know everyone else is thinking the same. Now we have reasonable doubts in something which relies on trust.
Basically, game theory. The system will find it's Nash equilibrium at a point where every individual move will worsen that individual's standing.
To break this spell, you need agreements and contracts. Someone needs to work on that, negotiate and lobby for it. But who? Would anyone who would benefit from that step away from their actual work and work on that meta-system instead? Would anyone who would not benefit from that system work on it? Maybe this could be a research project for scientists who already study these topics. Otherwise, I don't know.