When I previously ran into this situation, I simply changed the ErrorDocument directive to instead pull a PHP script that sent the usual HTTP 404 status code and HTML 404 error message along with helpful links pointing to useful sections and pages of the site. The script then also recorded all those accesses in a separate analytics profile. It was easy to see within a few weeks whether any of those links generated any meaningful traffic.
It was then an easy matter to change the few that were useful to redirect to real content, and carry on serving 404 to the rest. The data can be revisited from time to time. The data also showed a number of malformed links that were meant to point to real pages but where a typo had been introduced. Redirecting those requests added about 5% to the number of useful incoming links, several giving reasonable traffic numbers and previously unfulfilled due to the typos.