I've taken an entire day to think about it, and ADGarner, you do make some very good points. The metagame is roughly where it needs to be, and any changes that would mess it up should be avoided. We don't want to encourage players to break the Summoners Code because of a new system, and people shouldn't have to wait in a queue and get ready for a fun game, just to watch the other team surrender in the first three minutes just because they got first blood, and it broke the other teams heart. I see where you're coming from with your concern about the mercy rule system "protecting players from games that are going poorly." The goal with it, as I see it, is instead to protect players from games that are going poorly and dragging out for an obscene amount of time. While I can't devise a solution to properly differentiate the two, I've been thinking all day about ways to address the original issue, while attempting to sidestep many of the problems you have described.
All of the methods we have described so far have been aimed at forcing teammates with wildly different viewpoints on surrendering to all agree one way or the other. In some cases, that can happen, but not in many because in many cases, players decisions are based on their philosophy of play rather than what's actually going on on the field (This goes both ways; there are just as many, perhaps more, people guilty of throwing up surrender votes after the first few teamfights as there are people who just vote No as a rule.)
Instead of forcing players to agree on something through threats of IP loss or disconnects, wouldn't it be better to address the issue at its source? All Riot would have to do is make surrender votes a hidden tracked statistic (like non-ranked Elo), and modify the matchmaking system to queue people with similar voting patterns together more often. I know it sounds stupid, but think about it for a second. It would keep people from disagreeing as often about when to surrender by putting people who vote similarly all on the same team. People wouldn't get angry as frequently over the team surrendering too early or too late, because oftentimes the rest of their team is in the same boat they are.
This would (hopefully) avoid a metagame shift because technically the people are still voting the same way they always did. This way, they just agree more often. While it would be true that people who choose to surrender would, on average, win games slightly less frequently than the people who don't (because people who surrender would pull off successful surrenders more often at first, denying them the occasional comeback win), this would just re-adjust the Elo of said players until they are playing matches at a level where the victor would be ambiguous for longer, meaning that they agree to surrender only after playing the game for much longer (Note: Building teams with the intent to "spook" the other team into surrendering when they stand a good chance of winning would seem like a good strategy, but most likely would flop because, barring very unskilled players, most people can realize that if an enemy buys early game dead end items they probably won't scale well late game, giving them incentive to keep playing). Since the actual surrendering system isn't being modified, speculation wouldn't be forced and feeding wouldn't be encouraged because the stakes are exactly the same as they were before. Likewise, since the surrender system isn't being changed, players will still have to wait their twenty minutes before voting would be available.
I admit this system does have a few issues, some of which I've tried to come up with solutions for. They are as follows:
1. It might cause an abrupt increase in surrender frequency in the first few days, tapering off over time as people's Elos readjust, but always persisting to a minor degree.
2. It might cause an increase in time spent waiting to join a queue (because the system has to find people with comparable Elos and comparable voting patterns).
This one would diminish over time, though, as more players join the League and the system has a larger pool to choose from.
3.The math might get a bit complicated (because now the system has to consider two variables instead of one).
An efficient method of dealing with this would be to represent a players condition as a vector <a,b> with "a" being the players Elo, and "b" being a coefficient representing the players voting preference. The system could determine which players belong on the same team by matching players whose vectors have the highest dot products. This would mean that the matchmaking system would put more emphasis on teams that agree at lower Elo (where teams are more likely to break into an argument about surrendering or not), and less emphasis on it at higher Elos, where players are far more concerned about not waiting as long in queue. Dot products are really simple to program, and since it requires less code than other approaches, should have the least potential of messing up.
4. Riot could be doing other cool things for us with their time.
Well, there's no getting around that. But I imagine that sooner or later they're going to tinker with the matchmaking system anyway (or rather, the way Elo is stored or distributed), to combat Elo inflation. While they're at it, they could possibly use this suggestion as part of the new changes.
I'm sure there might be a few more problems, but it's a work in progress. Thank you for your input, it helped me greatly.
Do I still get that 1v1?
Edit: They call me the word mason because I build such big walls of text.