A key point for variation in quality when it comes to formatting, it seems to me, is recent vs. backlist titles. Recent titles, overall, tend to be better than backlist titles, which often seem to be hasty scan/OCR jobs with no proofreading.
Then again, last year's Harry Dresden book was a new release from a big publisher that had huge spaces between each paragraph and some horrible HTML underneath. I agree that the best way to know for sure is to check out a sample. Big, consistent formatting problems (like huge spaces between paragraphs) are usually apparent very quickly.