In no particular order. Design goals, best practices, aesthetic principles. Some things I find myself striving toward, talking about, doing. It's categorically untrue that I always do all of these things, of course. But maybe you can see what I'm reaching toward.
Feel free to add, elaborate, or dispute in comments.
Content is a reward. Make sure the content you provide is worth the effort you're rewarding. Even an autoresponder should be worth the time it takes to read it.
Never, never let effort go unrewarded, even if the effort isn't what you expected or wanted.
That said, don't excessively reward players heading in the wrong direction -- the volume with which you respond is an indication as to whether they're barking up the wrong tree.
Each discrete piece of your game should be independently entertaining in its own right, even if the player never sees another piece of it.
If you want to make a game for a mass audience, there should be something for every level of involvement, if possible, and for as many kinds of players as you can manage: explorers, achievers, socializers, killers; more than just spectators, speculators and solvers. Read up on Dr. Bartle.
Try to make an experience that would make sense even to a single player who is too shy or otherwise unmotivated to find or join a community.
Be aware of trolls. Consider interactive elements of your design from the perspective of somebody who has the most fun when defecating into somebody else's swimming pool.
Reference new content streams from within a known-in-game source as soon after discovery as possible. Once firmly established, this habit both supports the players' effort by acknowledging it and helps to prevent gamejacking.
Again, for a large-scale game, whenever possible, keep all communication in the open. Do as little as possible through private IM and email. This scales much better and will save you time and headaches.
Consider structure. Completely open-ended games can lose players who miss a road sign and get lost. Structure can mean a guide through the experience (often a cute brunette girl). It can mean a central website acting as a story clearinghouse. Just make sure players know where to look, and when, rather than guessing.
Make sure it's obvious to your players what their current goal is. Uncertainty isn't that much fun.
Use your structure to provide clear calls to action from time to time. It's helpful to get everybody on the same page now and again.
Provide rolling recaps. This serves two purposes: It allows existing players to attend to an urgency (vacation, term paper, conference) without risk of losing the thread of the story. It also allows new players a simple way to jump in, even late in the game.
Value your players' time and attention as much as they do. Don't release a lot of content solely for the sake of having a lot of content; don't create a lot of puzzles solely for the sake of having a lot of puzzles. It's surprisingly easy to overwhelm players with more information than they can process at once.
Manage expectations carefully. Don't commit yourself to a volume of content you can't realistically do, like live IM around the clock, or fifteen updates every single day.
The players will generally care less about plausibility than you do. Still, you need to put in the footwork on the motivations and actions of your cast. Make sure you understand how all of the parts of your game fit together, or the gears will grind instead of spinning.
Never let realism get in the way of fun.