In the previous thread, one of the commenters noted: Of course, OGH has previously estimated that disaggregating the publisher's job would land him with 0.5-equivalent of management work, leaving us (and him) with only 0.5-equivalent of an author. That doesn't mean it can't be done, but there'd have to be a pretty good reason...
Let me expand on that, in case anyone isn't convinced.
Left to my own devices, in a good year with no major disruptions (which, alas, don't come along as often as I'd like) I can write around 200-240,000 words of finished fiction — a pair of 330 page novels or one big doorstep plus a novella. This assumes I'm working on lightweight novels that flow easily, or that my drill sergeant muse is standing on my shoulder shouting "gimme chapters, worm!" in my ear through a megaphone. (A tough one can flow at half the speed — "Rule 34" took 18 months to write, for example.)
The specialist SF trade fiction publishers I know have a production ratio of roughly 6 novels/year for direct employed members of staff. That is: Baen (10 folks) produce 60-odd novels, Tor (50 folks) produce 300-odd books.
However a modern trade-fic publisher is an organization dedicated to handling the work-flow of book production. Over the past 30 years they've ruthlessly outsourced everything that isn't a core part of the job of publishing — including many tasks that an outsider might think were core competencies. Copy editors work freelance, paid by the book. Proofreaders ditto. Typesetting is carried out by DTP agencies. Printing is the job of a printer, not a publisher.
The stuff that remains in-house is editorial, marketing, accounting, and (occasionally) sales. "Editorial" in this context means workflow management — someone to ride herd on the pool of copy editors and proofreaders and to make acquisition decisions (in their spare time). "Marketing" includes book design, blurb writing, ARCs/review copies, presence at trade shows, glad-handing the big chain buyers, commissioning advertising, organizing signing tours and author promotion, and so on. (There's also a "production" side, sometimes subsumed under editorial, whose job it is to organize typesetting and printing and the business of turning the manuscript into a physical product. Generating ebooks slots into this workflow in place of "send PDF file to printer, order x thousand copies").
Part of the workflow involves running the copy-edited manuscript past the author for comment, and letting the author rub their bleeding eyeballs over the page proofs. This isn't part of the 6 books/year/employee, but goes on the author's side of the balance sheet in addition to "write the book in the first place".
So, I estimate a book takes roughly 2 months of publishing company employee labour to produce. Plus another 4 weeks of author eyeball time (which is that part of the publisher workflow the author undertakes — see previous paragraph).
When you add it all up: if I'm as efficient as a trade publisher, it would take me roughly 3 months to produce a book that also took me 6 months to write. More realistically, I'm likely to be less streamlined and efficient than a publisher who specializes in this job. This supposes I'm sufficiently plugged-in to commission my own copy-editor, book designer, cover artist, and typesetter. I then have to handle the contractual, accounting, and tax side of things. Ebooks are not VAT-exempt in the UK, so there's quarterly VAT accounting. Sales through Amazon.com outside the UK are liable for US tax withholding unless I jump through various international tax-treaty related hoops. HMRC are likely to sit up and pay attention if I suddenly switch from being a trade supplier to a direct seller, so I'd better have my books ready for inspection at any time.
I reckon 3 months of management time per self-published book is probably an optimistic, low-ball estimate. I should also note that, as a worker, my strength lies in generating crazy new ideas and hacking out new stories, not in sales accountancy, project management, and attention to fiddly details. (This is probably true for many if not most authors of fiction.)
Anyway: this is why I don't self-publish. Yes, I could do it. But it'd suck up a huge amount of time I would prefer to spend doing what I enjoy (writing) and force me to do stuff I do not enjoy (reading contracts, accounting, managing other people). The only sane way to do it would be to hire someone else to do all the boring crap on my behalf. And do you know what we call people who do that? We call them publishers.