I dont know how people are playing there, but I don't think it is much different than me. So, I tell about my experience. It takes dozens of hours to build a good fortress, then you get bored with it and start another. The former is lost or destroyed (anyway it is not a point of interest, i never look at it again). The community games end in a such way too (as I understand). I've seen 20 pages of fortresses which weren't updated for at least a month (and suppose amount of resources, e.g. time, was used to build it!).
It is possible to utilize results of playing in a such way: create a resource like dfma, where people would upload their fortresses. Then, when the world is generated, and fortresses of civs are placed on the map, the generator can use this resource, to search for matching fortresses and substitute random fortresses with peoples'. So, my thoughts:
0) You'll get superb locations for dwarf mode (current results of generation are poor compared to people's fortresses, and they would be poor) - fortresses, ruins, human settlements (I've seen, some people like to play it) and so on.
1) Matching is hard. You should match rock, soil and temperature. Hopefully, 20 pages of community fortresses should negate it. In a bad case, use appoximate matching.
2) Placing is a technical task (no new content is required), although it is large work: all you have to do, is to align local time counters and populate it with history mosters.
3) Bandwidth. My saves are about 10-15Mb now. Stripping them from all non-relevant info should help, but i dont believe it helps much. Anyway, the restriction could lead to "packs" of fortresses(hundreds of megabytes), which would be downloaded separately
... in game downloading would be better.
4) Versions and mods - another problem to keep an eye on, if we export only buildings it shouldn't bother much.
5) Moderation would be required - at least to classify the fortress: human town, dwarf fortress - architecture style; their size etc.