You're right about the engine of course. I tend to think in terms of, "Use toolset to do "X". Can't do it! Obviously need new toolset," but I'm neither a developer, nor a scripter/modder/etc. What I'm thinking is that it seems like the GNU-Linux-OpenSource community is doing a pretty bang up job in a lot of areas like OpenOffice and Gimp, and NWN shows what kind of support a game can get when the (good) tools get released... I know that writing an engine is very different to scripting and expanding with a prefab toolset, but just think:
What if we could come up with an engine that was somehow modular? Graphics, interface, rulesets, etc. It needn't be all things to all people if we start with a few assumptions, many of which were hit with NWN. They might be:
1) Seek to extend classic PnP. This suggests the NWN top-down approach, DM flexibility and the presence of a ruleset. Again, perhaps rulesets could be changed up or plugged in. Personally, I like how scripted encounters mean you don't necessarily NEED a DM, even if it's more fun to have one.
2) Multiplayer is the primary focus. The ability to create single player mods is awesome in Neverwinter. There are great stories to tell and questions to ask, but for me, that goes through the roof once it becomes social, hence the MMO cash spigots. And the more flexible the size of the player base can be, the better. The trade-offs will always have a lot to do with hardware and bandwidth limitations versus immersion, but wouldn't NWN be great if you could go with even more players online if you wanted to? Also, the ability to jump from server to server was genius (they really did get so much right), but how many of those limitations will become less problematic with innovations like cloud computing?
3) Immersion. This can be very subjective. For my part, one of the things I love about PnP is that my imagination has great graphics. On the other hand, they're limited to what's in my own head. I think Sephira was dead on about what graphics bring to NWN2. Sometimes a great picture is enough to greatly enhance immersion like in some of the old rpgs and MUDs, but then you have games like Oblivion where (story and solitude aside) you can lose yourself in the details. I work in visual effects, and we work a lot with proxies for images. When resources can't handle the shear volume of information as in, say, a 2k film plate, a proxy is useful for many operations where the detail doesn't need to be there. Is there something similar that could be implemented in our "modular" game engine? It needn't be Oblivion detail all the time, but perhaps there is some scaleability. It might only be in the textures and tilesets as is very common (I think NWN has this, right?), but it might extend to poly counts and beyond. Again, modules might bolt on like the CEP does now, adding flexibility.
Realistically, I know that characters wont be able to kick over a wooden table and take cover behind it any time soon. I can *twitch* accept that. But NWN was a huge step forward, and maybe it started with brainstorms like this. There are DEEP pools of talent in this community and elsewhere in open source and so on. What do you think?
Disclaimer: Like I say, I don't know much about the nuts and bolts it would take to do this, I'm just thinking out loud...on a message board... which isn't that loud, I guess. I may be completely off about what goes into creating a game engine, in which case, corrections are welcome.

Also, I've seen ORPG and played WebRPG back in the day. Way cool, but imho they are the infancy of what may be right now, with nothing but current technology.
Jadus