DOOM 3 engine will be scalable, just like all id engines in the past have been. So no, you don't need a uber PC to play the game. And Carmack's netcode has always been brilliant. Even DOOM's P2P play was revolutionarily fast for its time.
If I'm not wrong, DOOM 3 will be extremely moddable. Robert Duffy stated that almost all game-related files of DOOM (except for media files) can be open as a text file using notepad. All you need is to learn D3's scripting language if you want to edit any part of the game, or mod for it. Kinda like learning HTML and using it to create webpages via notepad.
Then you mentioned, "if there is so little emphasis on Multiplayer, what good is it going to present to the MOD community which help put Doom and Quake the hits/classics they are?"
If you could remember, the original DOOM only supports 4-player P2P MP. Yet even up till today, people are playing it for the sheer visceral experience. Heck, I've even replayed DOOM SP for nearly 20 times!
AFAIK, Carmack's choice for having P2P MP for DOOM 3 is because of technical limitations. The main culprit is D3's unified lighting system, in which light from all sources bears exactly the same properties and reacts the same way to its surroundings. Unlike every other game, the ULS in DOOM 3 rules out projected textures, light maps, shadowmaps, and all other lighting solution. Everything is generalised and all lighting and shadow calculations+rendering are to be handled in-engine. That's a far cry from the "dynamic lighting" produced by prerendered shadowmaps, as witnessed in games like Splinter Cell, Half-Life 2, or even Deus Ex 2 (though shadowmapping is used sparingly in DX2).
However, the brilliant visual experience produced by D3's ULS comes at a price, since to can't "generalised" the game environment using the client-server model. Sure you can do lighting estimations using client-server model, but that's just that. Estimations are not always accurate (due to different ping rates from different players), and not wanting to break the consistency of his lighting engine, Carmack decided to implement the P2P system for D3's MP.
IIRC, Carmack also stated that the physics engine in D3 will be a generalised solution for all ingame applications. Unlike other licensed physics engine, D3's will be fully integrated into the game's core source code, allowing the ingame physics and lighting to mesh seamlessly together. The precision of the physics model will also not allow for inaccuracies, especially in high speed MP DM, hence further attributing to Carmack's decision to solely implement P2P MP model.
Personally, I would rather sacrifice large MP player capacity for all the eye-candy and intense atmosphere of 4-player DMs. After all, I'm really tired of 32-players DM or TDM with noobs scoring high no. of "lucky" kills.
On to your last comment. DOOM 3 MP is gonna be scalable, believe it or not. Robert Duffy even hinted that a 8-player DM MOD can be easily achieved. With hardware advancements, you probably can have 16-player or even 32-player MP MOD in 2-3 years time. Id doesn't want to invest their time working on the MP portion because they expect ample participation from their fans. And what's wrong with SP MODs? They are the rearest and most valuable type of MODs in the community today. To me, a good SP MOD has longer shelf life (or disk life) than a good MP MOD, since the former is few and far between.
Finally, I would like to add that I'm very much in anticipation for DOOM 3's Single Player game. From the looks and sounds of it, it seems to be shaping up as a new System Shock, with better level designs, scarier and more bizarre monsters (demons), plus an exclusive trip to Hell!