It becomes more and more difficult to navigate within PO (especially in the review section) when one is looking for specific things but doesn't remember the exact title, and this will become a nightmare when as planned at some time in the future, the whole arXiv is imported for review.
I'd like to propose a new feature that would not only make this navigation very user-friendly but - existing nowhere else - would make the site unique and perhaps more attractive because of this feature. On the other hand, it will take some effort to get it implemented.
The proposed feature is an enhancement of the current possibility to mark papers as favorites. I'd like to propose that every user with a significant reputation ($\ge50$?) gets access to a structure that serves as a personal home directory (PHD) for maintaining one's PO related activities.
This PHD would just consist of a tree of folders whose leaves are symbolic links to arbitrary PO pages or (one's own) other folders. The naming of the folders and the symbolic links should be up to the user (with a default for links if no name is given), and allow at least alphanumerical strings (including blanks, apostrophs and commas) of length $\le 80$ characters. Pages looked at by the user but not yet referenced in the PHD should be placed automatically into a default folder ''seen''. Similarly, papers contributed to by the user but not yet referenced in the PHD should be placed automatically into a default folder ''contributed''. There should be facilities at least for listing, renaming, moving and removing folders and links, for copying single links for copying multiple links marked in a list, and time stamps indicating the last change.
In this way regularly participating users can organize their pages in a personal way which makes it easy to quickly find items to work on, and to remember their incomplete activities. Since only symbolic links are stored in a PHD, the amount of extra storage needed is minimal. If necessary one could limit the total directory size to a size roughly proportional to the maximal reputation a user had in the past.
With some additional implementation work (and of interest in the context of extending the review section) the PHD can even be enhanced to a system for managing the complete bibliographic information one is interested in. This would mean that any user with PHD privileges would have to be able to submit bibtex references (which would automatically become candidates for review), and there would have to be a facility for extracting bibtex files defined by the entries of any given folder.