As someone who has worked the developer side of the mod community (while I was at id), encouraging and aiding community members to enhance my employer's game with mods and extend its life, I think the nail has been hit on the head by a number of posts here. Here are my assessments of what is likely going on with poor PPM mod support for the game:
Poor multiplayer support in the game itself. id (in the person of John Carmack) admits that Quake 3 was to be their end-all and be-all of deathmatch style multiplayer. id made attempts to support the community, but it was all mostly unofficial ... and not smiled upon by id management. Team arena wasn't the greatest game expansion in the world, but it never should have been as complicated as it was to make your own team for that game. I still tip my hat to the Brothers Grimm and Dark Horizon for their work on their Tour de Force original team project.
Lackluster multiplayer. Maybe I run in the wrong circles these days, but I don't hear any groundlevel player buzz for starting up D3 servers. It has been my experience that people make player models and skins for games they love to play, so they can show off skills, personalize teams. Quake had that in spades; Quake 2 had some of the most original and creative player models I've seen for any game and no end of player skins ... and it had a mode for transferring that content to other players; and Quake 3 had a variety of interesting player models upon which to mount skins, but pure servers and the tendency of players to force player models on opponents (everyone became the glow-in-the-dark Bones character), and the three-part character model set-up put a damper on some creativity (while inspiring others).
The bar is too high. Doom 3 has an overly-complicated production pipeline. id underestimated the complexity of their production pipeline for D3 from day one. The complexity of process required to get normal-mapped model into the game can be overwhelming, even for professionals. It's far and away too complex for casual artists and game mod makers. Even with the export tools being a part of the game code, the bar is too high for anyone without near pro skills to jump and make anything they are willing to share. And I don't even want to dwell on the tools needed for the pipeline. If you do it the way id did, then you need both Lightwave (for modeling), and Maya (for animating) in your toolset. Though I understand that converters exist for other tools now.
Style. If your work isn't done in the id style it won't fit the world. Quake 1 was klunky enough to allow a wide range of art styles ... it wasn't quite so embarrassing for a neophyte artist to share his art. Quake 2 (and UnReal) started raising the bar ... and the community was more than willing to accept a mix of cartoon characters, superheroes, and grim figures mixing it up together. Quake 3 and UT pretty much demanded that the art at least be polished for a model or skin to be accepted. But I just don't see that in Doom 3 ... or any of the super realistic game mods coming out right now.
Anyone who can make decent model art for a D3 PPM can work as an artist in the game industry ... if they desire. And for most folks who invest a lot of time in this process, that's their dream and goal.
Finally ... game diversity is with us now. In the "old days" everyone focused their multiplayer and mod activity on one hot game at time and the shelf life interest for that game was often years, not months. By the time Quake 2 shipped, it wasn't long before it was competing with Unreal for mod attention. Artists shifted their attention between Quake 3 and UT for several years. Now, there's an annual UT update and a host of aging Q3 engine games to distract modelers, and I won't even go into HL2 or BF'42 and its heirs.
The focus is gone.
Maybe id learned from Quake 3 and Team Arena ... if it's too easy for amateur game makers to make mod content for your game, it will be difficult to convince the public to buy your expansion packs.