Just hopped in IRC for a few minutes and posited a similar (but for more more brief) version of the idea.
It seems that the all or nothing aspect is really a lot more offputting than I had assumed it would be.
Turns out, that working for an entire week and not getting paid for it, perhaps as often as 4 out of every 5 tries, could really be a problem from a practical standpoint.
I had assumed that if you were really that bad, you would look to improve really quick or perhaps only bother to bid on something you have a great deal more skill in.
Money is a great motivator for me and lack of money has always been an fearsome teacher.
I guess most people see it differently.
As it happens though the root cause of some of the "gaming the system" problems mentioned earlier directly stems from the lone wolf.
Teams are the best defense without resorting to stupidly draconian methods such as limiting based on IP address (not possible in most college settings and even many cell carriers are giving everyone the same IP)
I think I'm really just opposed to the concept of long term persistent teams
seriously I'm a freelancer, if wanted to be part of a long lived team I'd go get a real job.
Even worse is automatically creating a team from randomly selected mates.
There is the real risk of the team members not even being able to feel out roles before the milestone is over.
We had a class in "team building' once that had a similar problem, the team members were assigned by random draw and in a week the entire team had to come up with a presentation and present it to everyone else. The end result was not good. As they say "too many chefs spoil the soup".
Maybe it doesn't have to be one or the other, perhaps there are other possibilities not previously considered.
The first which I'm considering, is based on a game I played years ago and resulted in some really good teams. (and many real projects narrowly avoiding slipped delivery dates)
A team was comprised of a leader/coordinator, and 4 or more functionary members. I say functionary because each member is given a specific role based upon their experience and capabilities. In the game, a team leader was just the first person who decided "Hey I think I'll start a team". That person would then select the roles they needed to have filled. For instance (drawing from that same popular MMO) a team would be comprised of a Tank, a Healer, a Direct Damage Dealer, and someone with stealth capabilities. The tank would take the MOB on 1 to one to keep it occupied, the DDD would nuke the MOB to bring it down quickly, the stealth guy would run around opening chests and grabbing loot, while the healer would try to keep everyone alive.
Generally these are pickup groups and they would form, go on a mission or two and split up.
While we aren't talking about crawling dungeons and raiding treasure chests, there are some striking similarities.
What makes a leader capable of leading is the desire to lead and the confidence to pull it off. To be a leader you must know what you need in team members and if we added a "social" aspect to the site. It shouldn't be much harder to create a team on the site, than it was in the aforementioned game.
The team leader from each team could bid and a single team would be selected at random during the bid process instead of 5 individuals.
The team as whole is given 1 week to pull off the milestone. The milestone payment is then automatically split amongst the team members at the end of the milestone according to the method that all members agreed upon before joining the team. (in otherwords, the split could be varied, but you would be guaranteed the split you were offered).
The biggest difference here, is that there is only a single team selected, there is only a single target delivered (instead of 5), 360 feedback from other team members is given at the end, and if anything at all is delivered, then payment becomes mandatory. Everyone gets paid. We may need a community dispute resolution team yada, yada, yada...
To prevent perma-teams, a team would exist for a single milestone and then be split up. They would need to go a fixed amount of time (perhaps a month or so) before working with any of the others again.
By not allowing the same 5 people to continue in a team beyond the space of a single milestone, it should prevent much of the gaming of the system. It would also promote co-operation and socializing which may turn out to be an important aspect in terms of keeping eyeballs on the site.