In computer science these days, there are very few journals that do any of this for you. The standard, even for expensive, non-open-access Springer journals, is to require the authors to submit camera-ready PDFs. Maybe that isn't ideal, but the closed-access journals mostly aren't any more full-service than the open-access journals on that front, so it's not clear what the money is buying us as authors. All the publishers provide are Word and LaTeX templates, some formatting instructions, and loud complaints if you get anything wrong.
The only major journals in my area I can think of that do significant in-house professional layout are the IEEE's "magazine-style" journals targeted at a mixed academic/professional audience (they go for a full-color, glossy-magazine look). The IEEE journals that don't fall into that category require authors to submit camera-ready PDFs, though.
Quality always costs. I think that open access journals are net cheaper than their print counterparts, but it will always cost money to produce a good journal. Look at the PLoS group. They are the closest we have to a top-tier online-only journal and they produce a high quality journal with an excellent impact score. However, it still costs $2900 to get an article published (if it's accepted) in PLoS Biology. Why? Because that's their costs.
You are also severely over estimating the ability of most academics to produce a good, publication ready, PDF. If you're publishing in a journal, there is a set look and feel that must be applied. No matter how good the work is, someone still has to apply the right layouts. I've published in a conference before where what you submit is a camera-ready document, but somehow no one seems to set things up the same way. And good luck getting a biologist to figure out LaTex... that isn't going to happen.
What you're describing seems to be closer to a distributed combination of arXiv and PLoS One, which is where I think we'll end up moving in the future. However, for now, the hosting costs of the organization is not trivial. Even arXiv has started seeking external funding for support.
I'm scared to think of the state of science if open journals are going to do to existing journals what craigslist did to newspapers. Now newspapers are struggling to figure out a way to pay for high-quality reporting in a world without a sustainable revenue model. The result is poorer quality news or news dominated by a few big players.
In science, the journal system has long served as a filter for information. Sure, a purely open access environment would let more people publish quicker, but it would also fill up the scientific record with a lot more noise. And there will always be costs to filtering signal from noise.