This could apply to more formats than epub, but epub is all I really care about it for.
There are a number of types of epub formatting that aren't necessarily easily be fixed programmatically, but should be easy to detect programmatically:
- Justification - any book missing 'text-align:justify' in it's css is probably left aligned
- Font Family - any ebook missing '\bserif' (regex) in the css is probably sans-serif
- Margins on body or @page - can be fixed programatically as discussed in the Modify epub thread
There are probably similar types of checks on the css that could be done, but these are common ones I find myself detecting and fixing manually all the time.