The idea sounds mostly cool, however, I'm seeing a few possible stumbling blocks.
1. (The minor thing):
Rolling by version and stuff throws the probability way off. In effect, it favors characters from smaller games over those from larger ones. As, say, Evo has the same odds of getting picked as V4, that means the chances for selecting one out of twenty-one characters and the chances of selecting one out of two hundred and seventy-six are the same. It's not even. The alternative is to throw all the characters into one huge list and just roll straight by character. That, however, favors certain versions (Evo will never get picked, then). So... the question is, what's more important here? Wide exposure to versions, or to characters? I don't have an answer; I just want to bring up the math.
2. (The more major thing):
There are some handlers who wish certain characters of theirs would be forgotten, and some characters who... are not going to get a lot of positive feedback. My worry is there's the potential for this to end up as open season on those who are "acceptable targets" (which we definitely do have). I guess this goes with the "Don't be a jerk" rule, but maybe just note that criticism, likes and dislikes, etc. should be phrased constructively and politely? Basically, if you roll, say, Blood Boy, even though he's not so popular these days, I feel like everyone should be encouraged to not get mean about stuff.
3. (The last thing):
I far prefer this as user-run than as staff-run. It's awesome to see the members getting involved in stuff, and I think this works better as an unofficial sort of thing.
That's basically all. You've also given me another idea, but I'll throw it up in a separate thread.