See here and here if you wish to know about Polymath.
I for one would love to be a part of some collaborative problem solving endeavor. Although it can be said that quite some physics research is already massively collaborative -the LHC- for example, but I am talking about some research level theoretical question. There are some issues that I thought about, and I would like to here your own:
Lots of people are already working on lots of problems individually, for their thesis etc. An ideal question should be somewhere between an examination question and a thesis report for such a project to be feasible. (i.e too tough for an exam, but not big/exhaustive enough for an entire thesis)
Some senior practicing physicist(s) must mentor/administer such a project. This is the only way to ensure that nonsensical things do not get accepted by some sort of crowd consensus.
How is it different from physics.SE? This is where I have my greatest apprehension. Here people post their own solutions which are intended to be complete. These are then judged and voted accordingly as per their accuracy and clarity, etc. I don't see how this can be applied to physics problems, although it might be possible. For example, a Maths problem the collaborative effort usually goes like this.. people start by testing for trivial cases, trying induction, trying to reduce the problem to an easier or known one, etc. In physics, if we are to study some physical problem, we might have to make simplifications to put it into an analyzable-framework. It then reduces to using known physical laws and applying it to the specifics of the problem.
I don't know how clear I have been, I would like to know what you all think about this. I am in no position qualified to start such a project, but I would have loved to participate if such a thing existed.
Notes: from the comments
Need a blog or an equivalent platform with commenting-type feature as the place to release problems to general public. Will anyone volunteer to host such a problem in his/her blog/website?
phy.SE might be the best place to propose problems. I beleive a problem has to be good enough to allow different kinds of approaches so that we may learn from different insights. However, I guess some problem proposals here, even sketchy, might be a good idea to start.