Cool to hear that this is your area of interest! This is precisely the sort of thing I do research on; I'll try to keep my thoughts brief.

How are traditional face-to-face (FTF) communication processes different from computer-mediated communication (CMC) processes? Well, it used to be the case that one could say that CMC was synonymous with text-based interaction. In this way we could sketch out the differences between text-based interaction and face-to-face interaction. No longer is this the case, as we all know now, there are pictures and videos and other stuff online. Nonetheless, one truism that is repeatedly found in the scientific literature about CMC vs. FTF interaction is that forming impressions, entering into social relationships, becoming relational, and so forth all takes longer via the Internet than it tends to in FTF interaction. (See Walther, 1992; Walther & Burgoon, 1992; Ramirez, 2007).

Another interesting feature of CMC is that in many cases it tends to be more "social" and positive than anything else (Walther, 1996; Walther, 2007). Now there are competing theoretical accounts--hyperpersonal communication (Walther, 1996) on one hand, and the social identity model of deindividuation effects (SIDE; Reicher, Spears, & Postmes, 1995) on the other. Hyperpersonal communication argues that because of we can selectively self-present, idealize our partner, and become more focused upon messages that there's not as much time constraint to processing we actually find the web to be a more "social" environment than we do FTF interactions. The SIDE model argues that we stop seeing others as actual people and start seeing them, primarily, as group members. The more these group members adhere to social norms consistent with the group, the more we like them. Because this can happen to a greater degree online, the SIDE model argues that this occurs more prevalently on the Internet.

I don't buy the SIDE argument for a number of theoretical reasons that I won't expound upon here. To be fair, the biggest reason I don't buy it is that my doctoral advisor was a (the?) primary figure in the opposition to the SIDE model. Either way, we tend to find that the web is an exceptionally social place.

One thing that, if you're talking about Downfall anyway, you might want to talk about is the application process here. Which, from a social science perspective might be the most interesting thing we do as a guild. We tend to take a great deal of time before we're willing to let another person into the fray. According to Social Information Processing Theory (Walther, 1992) this makes a great deal of sense. It's also consistent with the idea that online gamers are primarily motivated not by game achievement, but rather by social connectedness which is available online (Carr & Van Der Heide, 2009).

I'm not sure what you mean when you say that Downfall has created something beyond a virtual community. It's certainly not ONLY virtual if that's what you mean (i.e. from it's beginnings as a group of RL friends, guild meet-ups, etc), I'm not sure that it's something more successful than other virtual communities from a scientific perspective. (Don't get me wrong I think DF is awesome; I'm just saying that the perspective that one's virtual communities is / are awesome is a perspective that isn't all that uncommon, according to the data which support the hyperpersonal model.)

Nonetheless I think your observations about DF are fascinating. Here's why I think people here think DF is great and they enjoy their friends from there so much.

1. We've all had an extended period of time wherein to create our online personas for a fairly regular audience. That is, we've been offered almost endless opportunity to selectively-self present. The "me" that you all see is the one that I've carefully edited and chosen to give to you. (Aside: Does that mean non-selective emo-outs never happen? Of course not. Many of them are even quite memorable. It's just that when you consider that you'd likely have emo-outs in FTF communities too the channels (verbal and nonverbal alike) are much broader. When people become frustrated they can roll their eyes. In G-chat, there's no real amount of eyerolling that can successfully convey that a person is pissed. Thus, they move to the direct verbal emo-out to express frustration. It's not so much that people are bigger dicks online, it's that their usual channels for being a dick are restrained--thus, they resort to verbal dickery.)

2. We've all had an extended period of time to come to believe that people are the best possible version of themselves. Hey we all want to make friends, right? There's a general predisposition to present oneself positively (Goffman, 1959) and there's a general predisposition to think of others positively. When given the opportunity to idealize others, we are only all too happy to idealize away.

3. The channel (CMC) allows us to focus more intently on messages' content rather than the packaging a message arrives in. Because my messages on, say, this website can be thought about and ruminated upon carefully, one is more likely to engage the actual message than if we were having a FTF discussion about something.

4. A feedback loop exists. This is a difficult concept, but I think it's an important one too. A bunch of years ago three researchers (Snyder, Tanke, & Berscheid, 1976) did a study where they let two naive participants interact with one another--a boy and a girl. They showed the boy a picture of the girl that was randomly either a) hot or b) ugly. In behavioral coding of the interactions here's what they found. The girls who talked with boys who had been shown the hot pictures actually behaved in such a way that coders blind to study hypotheses and study conditions actually thought that the girls behaved more "hotly" and the girls who talked with boys who had been shown the ugly pictures actually behaved more "ug-ily". This is called behavioral confirmation. Read that finding again...it's a wierd one. It turns out that my beliefs about you can actually impact the ways that you interact. If I think you're hot, I'll look for confirmatory evidence and seek to give you the opportunity to confirm my suspicions. I think this occurs online. Because we all present the best possible version of ourselves to begin with, our partners (friends) idealize us and consequently give us further opportunities to confirm their positive expectations. Because of this, over time, we start to like people more than we otherwise would in Social Network environments.

This is way more than you wanted / asked for, sorry about that. Hope some of it helped.

A few of the things I'm thinking about including are the Internet Dickwad Theory and the story of "Talhotblond".