Its because most PIs could care less about the actual research, but rather only care about their 'rank' in the pissing contest that is academia. The internet should have, and in some more lucrative and distributed areas of the ivory tower it actually has like Archiv, made journals redundant and pointless. Only peer review should have remained. One would naively think this based on the idea that the academy is about knowledge. However, PIs can commonly be vain, narcissistic, and only promoted based on previous accomplishments. These are the people that got through via the random sieve that is modern schooling with nothing but A+s from kindergarten through grad school; they only know rankings and nothing else. As such, a barrier of luck under the guise of effort must then again be invented, this being the journal process. The reason is that there is not enough money for all the grad students out there to become PIS. Publish or perish is not a bug, but a feature that keeps out people from un-well connected labs and PIs. Remember, ~50% of all papers are never read (0) and ~50% of all papers have such large issues with their stats, that they can be considered noise (1). That means ~25% of papers are even relevant to the world in any meaningful way. Couple this with the average time for papers to appear from the submission to 'print' is ever increasing for 'new' authors (2), and you have what could be reasonably considered to be a concerted effort to keep out competitors.

(0) http://www.smithsonianmag.com/smart-news/half-academic-studi...

(1) https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1182327/

(2) https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4533968/

http://www.nature.com/news/the-mathematics-of-science-s-brok... (good overview of the issues)