There main reason is History.

Scientific research didn't begin in the era of the internet. Back in the olden days scientists basically just wrote letters to each other bragging up their latest discoveries. As the number of scientists grew, everyone writing letters to everyone else became too much work, so some folks decided to create a kind of magazine where scientists could write up their discoveries once-and-for-all to get word out to all of the magazine's readers at once, and the idea of the academic journal was born.

Like any good magazine, there were editors-in-chief who decided whether to accept or deny submissions based on if the article was sufficiently on-topic for their magazine's audience and whether the article was sufficiently cool to their audience. So some of those magazines became more famous as reliably having the best and coolest stuff in their respective fields.

Also, because it would cost a fair amount of money to type up all of these submissions, bind them together nicely into a booklet, and distribute them out to the subscribers, this was not done as a free service but rather - like most any other kind of magazine - it was done as a for-profit business.

The system wasn't perfect, but at the very least it was financially sustainable enough (for most of the best journals) to survive to the present day.

In modern times, the internet offers a multitude of ways to disseminate these same sorts of scientific write-ups to vastly more people at drastically lower costs (compared to oldschool paper-based printing and publishing). However, the historical system still has significant momentum due to the best journals still have the biggest audiences and a fair amount of prestige. This leaves academics with a tough decision when it comes time to publish: "Do I publish in a prestigious oldschool journal behind paywalls, or do I publish in some online upstart journal that is free-to-all?".

There are trade-offs to choosing either option. By publishing in a prestigious oldschool journal you are making your work harder to access (essentially limiting it to folks with academic library access) but you also get some bragging rights and free advertising based solely on the fact that your article is published by that particular prestigious journal. On the other hand, publishing in some upstart online journal means that anyone who clicks the link can read your work... but the low notoriety is of no help in attracting people to your paper in the first place.

Ideally, the science should stand for itself regardless of which avenue it is published under, but unfortunately the people doing the science have careers to consider in order to keep doing that science in the first place. So, whether we like it or not, the upfront bragging rights from a prestigious journal publication may be more beneficial to a researcher's career than a free-to-all online journal publication (even if that would be better for the scientific community). Ultimately, this means that publishing in upstart journals is basically a luxury that only already-well-established researchers can risk.