I had a try with a basic text, with one image and four footnotes.
I looked at the result on Sigil.We are not there yet.
+ the text is split according to TOC
+ a unique external and orderly stylesheet is created
+ processing is ultra-quick
- the text is split in random order
- the image is detected but not processed.
- the footnotes are detected. However, they are processed one by one, each footnote adding a split text part -in random order too-, instead of being put all together in a single footnote text part at the end of the book.
- the stylesheet has probably too many occurences of Georgia Garamond.
These are for the time being the main points.