My use-case: I download material in various formats from online, mostly in PDF, ePub, or some markup format (LaTeX, Markdown, HTML, etc.) I've got a large set of downloads, which I then try to import into Calibre. This is in support of a large research project.
1. It's difficult to tell what I've imported and what I haven't.
2. The import process itself is slow. Enough so that I'll fire it up, get caught up in other stuff, and ... well, tend not to get back to it.
3. The corpus is fairly large: around 1000 books and papers, plus another 5,000 others pulled from web archives.
4. Tracking this by metadata is crucial. Title, author, publication date, and tags. Managing _that_ is a headache on its own, especially adding metadata to works / confirming automatically extracted content is accurate.
5. Once I've got the information organized, reading, referencing, annotating, and other tasks should be supported.
Again: calibre is about the only tool out there I'm familiar with, but it's a pain. Zotero and various LaTeX bibliographic tools are also of some use.