Zotero is very good for several things, but it is quite lacking where annotation and knowledge organization is concerned. I suggest adding basic annotation handling functionality to allow it to better compete with apps that already include it (eg. Citavi, Mendeley, Papers, Qiqqa, and the now defunct Docear). Most of these allow annotations using a built-in PDF reader, but this isn't necessary if a program can extract and export annotations transparently. I propose the following:
1. Incorporating Zotfile's annotation extraction functionality into Zotero, and creating similar functionality for HTML and common office suite formats like *.docx and *.odt (possibly using Pandoc filters?).
2. Adding an "annotations" pane to the sidebar (alongside the "notes") and an "annotations" attachment type.
3. After opening and closing a file through Zotero, change dates and/or hashes are compared and the file is automatically re-scanned for annotations, and the pane and related attachment/s updated accordingly.
4. Each annotation is stored as a separate annotation-type attachment.
4.1. Attachments should (appear to) keep in sync with the original file, such that a change or deletion of an annotation results in a parallel change to an attachment. Deleted attachments are moved to the trash; how restoration is handled depend on the type of access Zotero has to the file (see below).
4.2. Zotero should have read/write access to annotations, but whether and how to implement it at this point remains TBD.
4.3. Attachments should be linked to the original file, such that clicking on an attachment will open the file at the correct position.
5. Zotero will provide a "view annotations" toggle (through a toolbar button and/or keyboard shortcut), that would set the main view to display only annotations, with/out their parent references. Similarly, annotations could be filtered out of view. Whether this view can be further sorted (eg. drag-and-drop annotations to reorder) remains TBD at this point.
6. Zotero will allow exporting annotations through the usual mechanisms: File menu, drag and drop, and word processor integration; including setting the inline citation format. The latter will be linked back to Zotero like any other citation.

This minimal set of functions would allow Zotero to be effectively, if not completely, used to organize reading and research, without making radical changes to the program or the way it is already used.