I remember I tested Lemon-8 once about two years ago - and probably it's still on my localhost-test server somewhere. It's a very interesting, yet extremely difficult, approach to start out making guesses about the meaning inherent in a document's structure. And that approach's certainly the most low-level and a good help for people who feel difficult to separate form and content of documents.
However I see you also understand why so many people choose a different approach, trying to apply some formatting constraints right in the source so the cleaning up work would be easier later. But the Lemon-8 type of approaches is not dead. It reminds me of the businesses built around conversions from PDF. I have not seen any totally conving result yet, but I've seen a lot of people / software trying to get something "meaningful" out of (non-tagged) PDF, and be it only something like XHTML or EPUB. It may even work well for fiction books, but it's very hard for academic / professional literature.
Anyway, I'll keep in mind that you're interested in those workflow topics. For the moment I'm very very occupied with customizing our new OJS-based website www.reinhardt-journals.de
, but I hope to get some more time in October / November to push our publishing workflows. If I encounter anything that might be of interest here, I'll let you know.
P.S.: Thanks for pointing me to oxgarage, I had never heard of this project before.