[Quoth]

I don't think any Java knowledge needed to run epubcheck?

@Pulpmeister, while I'd agree on the link attribute and chapter splits, those are minor and not generally a major issue. Easily avoided by doing nothing!

I import a single docx into Calibre and it seems to create an internal HTML file per chapter, at least in the epub2 conversion.

I used intermediate HTML years ago with Mobi Creator and Word 2002 /XP. No need with docx from Word2007 or later or Save As docx in LO Writer (but only open/edit the native odt, only save as docx for exports, as opening a docx does a conversion). Also an "anchor" on every line that's to be in the TOC and make those have a heading level. Any heading NOT in the TOC is then set to body text level, even if it looks like a heading. I auto create the Index by Insert Index (but page numbers deleted). I cut and past that to a text editor and paste back. Then give a Contents style and each line is then link #anchor.
On Calibre I check that the system epub TOC (NCX) generated then matches the manually built contents page that has links to anchors.

I test three kinds of Kindle with dual mobi (KF7/KF8) made from the epub2. The idiot DXG (only sees KF7), Kindle Keyboard / KK3 which can see the KF8 or KF7 and has more fonts as well as Publisher Font KF8 option and a PW3, which doesn't render KF8 quite the same.

I test the epub2 on Lithium, Kobo, Nook, Sony and Binatone if any style edit or new style, otherwise just on the Kobo Libra.

The key seems to be COMPATIBLE styles. If I make a new style or edit one I create a small document and load the docx into Calibre. I make an epub2 and view on Calibre viewer as the first check.

I'm lazy, so though I have 20+ years HTML and 15 years CSS experience and 40 years programming experience, the goal is to avoid editing CSS or HTML. Just the wordprocessor file.