Some more thoughts on a reboot. Looking at the list of successful (read: released) LMATs, they can be broken into two camps. Prismatic Palace and Omen Woods both had significant amounts of work done by a single primary organizer, either during or after the collaboration stage. This allowed them to be somewhat complicated in various ways. Aftermath and Kansas were much more varied in content, being a more direct result of many different people working together, but got away with it probably in large part due to the simpler art style (8-bit) and perspective (non-isometric). While both 8-bit and 16-bit art styles may work for individual authors, 16-bit is less practical in an LMAT setting because it severely restricts who may edit individual parts of the tileset. As for non-isometric perspective, I think that's a less important factor because anyone can still draw it either way, but it is both a) easier to draw and b) easier to add more stuff to because every solid surface requires fewer tiles that way. The more complicated the basic ground gets, the more someone is required to step in and do some serious independent organization work if nothing else to make sure everything snaps together.

It may sound like I'm leaning heavily towards suggesting the Kansas/Aftermath approach over the primary organizer/approach, and I am. Primary organizers can happen by chance, but I suspect that intending to have one from the get-go would make it less "Let's Make A Tileset" and more "Let's All Help So-and-So Make A Tileset," which feels less appealing somehow.

(I don't want to discount the importance of theme, but I think that's been talked about these threads a lot before, so I don't have much to add. See perhaps September for one example of that going wrong. Also the drawing style and perspective stuff can definitely be simplified too far: note that nobody has any really strong interest in releasing Valentine's Day, even though it's been essentially finished for years.)