If you have less points than other people, you don't get a grant, or you're more likely to not have your contract extended etc.
The ministry reevaluates the (journal's IF<->points per publication) mapping each year, but it takes a few years for a new journal to get significant number of points. Publishing in low-IF journals if your paper could be accepted by a major one is a career suicide.
The hidden implication is that the system reinforces the power of the all-powerful top journals.
So, in a lot of ways, the competitive research culture isn't helping inaccessible/exclusivity journal problem, since researchers are actively seeking out those journals.
Mathematics seems like the most promising field to start with if you follow a bowling pin type of strategy. There have been some boycotts of Elsevier etc. from math departments and they have somewhat of a "free publishing" tradition at least from my outside point of view. It's also an excellent viral starting point because you can infiltrate other disciplines that tend to cite math papers (physics, machine learning, economics etc.)
I doubt a system like the journals would have developed in a world where internet already existed.
I think a variation of the system used to build the Linux kernel could have worked for scientific papers as well. In the Kernel you have a pyramid of trust. Linux trust some people directly, who again each trust another set of people. This continues downwards. In this fashion code can be signed off and bubble up through the hierarchy.
I suspect in similar fashion a scientific paper could bubble up by being reviewed by ever more respected scientists in a hierarchy.