I was on the Stanford team (we sucked badly), and I must say it is the most frustrating thing to have a solution which works for every testcase you can think of, but is still rejected by the judges. And very demoralizing to watch eventual winner St. Petersburg State University sitting at the table next to yours and racking up the balloons (one for each problem solved).

One coach at the contest noted that, for U.S. teams, the teams are often formed by students having a local contest and just picking the students who do the best. The students then go off and find someone to be their "coach" for registration purposes only. Whereas for non-U.S. teams, there is usually a prof who is very active as coach, who coordinates team practices and such, and sometimes hand-picks the team members. Not sure how much of an effect this has on team performance.