We already have both systems running in parallel. Many researchers post PDFs immediately. Then they go through the process of trying to get them into a journal. Both have advantages. While the peer review and typesetting process of publishing in a journal has many issues, the changes generally make the papers better. It's usually easier to read the final published version than an early preprint.

It's not just inertia that keeps scientists using the old journals. Anyone can issue a preprint. The material that ends up in the journal is usually the better material. The filtering helps and so does the editing. The filtering may be flawed, but it's better than nothing.