Interesting. I wonder however whether all problem solvers contribute to the problems' rank or only the premium members. In case of the latter the number of solvers is probably too small to give a valid rank to the problems. In case of the former the ranks of the problems' rank will probably drift to one side (very weak or very strong) since the solvers often have to start at 30k.
i thought about this problem too. my solution is to have all members contribute, but to weight the significance of an event based on the size of the previous sample data. in other words, until a player has stabilized their rating, they won't affect the rating of a problem much, and vice versa. (this type of design is pretty common in rating systems btw).
in fact, even unregistered players on the site can get a stable ranking saved in a cookie as they progress, but i don't think i use this since we have enough data without it.