I wrote this last night, but decided to scrap it; however, it has become relevant to the discussion:
One of the biggest lessons BG has learned when it comes to infighting is 1) Update ideas 2) Factual information. The first is very discernible following forum changes. We previously separated Advanced and Standard. Advanced contained news information and "important" discussion. Standard was pretty much what we now call "Randomest Question Thread" (Better know a thread). Since standard kept producing better results as far as testing went, they decided to merge both forums. However, that did not end major arguments. It was not until we separated news information from advanced-standard that we saw a noticeable decline in major arguments. The ones that did remain were primarily fights over game data (Formulas and such), in-game rivalries displaced onto the forums, and moral issues (Windower/Jumping to WoE Fights/Other stuff). If we were to rearrange these, you'll see that the order of "resolvability" is as follows:
1. Game data
2. News information (Update ideas)
3. In-game rivalries
4. Moral issues
The last three have the most in common. Particularly, because they have no, or every feeble, objective criterion for evaluation. Without this objective criterion, you cannot end these debates. This is what you call politics. I know Japan has primarily been locked down by the LDP and Keiretsu + Gov = Mucho good times, so SE doesn't witness this, but politics has no end precisely because what constitutes the end is not objectively defined. News information is the least "non-objective" of these because there is a general progression towards wanting to improve the class and balance. Nevertheless, the level of balance (And whether it even exists) and mechanism of balance leave room for discontent. Game data has a very objective criterion: To learn the mechanics of the game. Assuming this isn't shifting every hour (i.e. a very reasonable assumption for those who are incapable of distinguishing assumption from its negative connotation), we have a general sense of what we want to do. We want to predict numbers before they occur in the game. Hence, debates are resolved via accuracy. At the same time, that is not always the case, but it is not because there is no objective criteria, it is because of the poster themselves. The next step is to ask: How do I discern the individual deviating from this objective criterion and fueling a pointless argument? That's going to be specific to the issue and unfortunately, that means you either need to hope all participants are "rational" (Doesn't actually exist, but it works for explaining it to you) or you need to be able to tell which posters are just flaunting posts. An example of the former is Rambus in the Elemental/Rational/Logic thread posted earlier. Despite Byrthnoth showing that Rambus's order of operations was incorrect, Rambus continued to spout absolute nonsense. This is not something that you can resolve via rule of the majority either. Most people don't know the mechanics and the report button becomes a "Vote for your favorite!" mechanic. It takes expertise to solve these issues.
In short, you'll need to lock or displace 3 and 4. Since you're looking for game ideas, you need to moderate 2 less than you have been. Modding the first takes expertise in the game or you should let the situation moderate itself.