I find the nacelle question is interesting.
But instead of walking into the conversation with my assumptions and expectations, I'd like to summarize what's already been argued.
Possible pros of more nacelles:
-- Higher maximum speed. Any warp material should have a power limit, a point at which more power will not induce a deeper warp bubble but will simply melt or otherwise damage the material. So, by having more, disconnected nacelles, you can put more power into the warp bubble without melting the coils. And though this same result could probably be achieved from larger coils and fewer nacelles, you can get faster results by using modular nacelles that are already designed, tested and known to be reliable.
-- Better manoeuvrability at warp. More total coil area may mean more complete control over the shape of the warp field and, thus, more and faster control of your direction of travel.
-- More "torque". Perhaps pouring more power into different sets of coils can give a warp field better "traction" within subspace, allowing a vessel to tow more mass at a given warp factor more efficiently.
-- More endurance. Less power per nacelle might mean the warp coils could sustain higher warp factors for longer periods of time.
-- Less wear and tear. There's quite a bit of energy impinging on the warp coils and that has to create stress. By alternately using different coils, you'll put less stress per cochrane-hour on each nacelle and they'll last longer.
-- Redundancy. The more nacelles there are, the more of them can be destroyed in a fight or a accident and still allow the ship to travel at warp.
-- More stable warp field. Perhaps having more warp coils keeping a warp bubble "inflated" makes the field less prone to interference to outside subspace instabilities.
Possible cons of more nacelles:
-- Hardware complexity. More nacelles mean more conduits, more junctions, more internal sensors, more hardware in general. It's gonna be hard to keep all that stuff working well together. And the more the hardware there is, the more there is to fail. Furthermore, the more complex the connections, the more likely something's gonna fail.
-- Maintenance complexity. As above, the more hardware there is, the more likely it is to fail, which means the more maintenance is needed to keep it in working order.
-- Software complexity. Starships are controlled by their computers. Computers need to be programmed. The more complex the program, the more bugs there are and the harder it is to find, and crush those bugs.
-- Warp plasma phasing complexity. The activation timing of warp coil segments by warp plasma both within a nacelle and between nacelles must be rather important. More nacelles means keeping track of more injectors and may well require both more accurate and more precise timing of coil activations than fewer coils. Furthermore, it's possible the state of the plasma is also important, making things even more complex.
-- More mass. Nacelles seem to be rather dense, and thus, rather massive. More nacelles means more mass... A lot more mass. In STL, more mass means slower response and less total achievable speed. This may also be true at FTL.
-- Poor economics of atoms. Four nacelles requires more warp coil material than two, and may require more dilithium to run. How many ships do you want to build with the materials at hand?
-- Poor economics of energy. It seems likely (via the "no free lunch" truism) that even if it takes less power per nacelle to maintain a warp field of any given factor, it will take more *total* power with all nacelles active. Further, any nacelle that is not powered is dead weight. This means more power per light-year and a less energy efficient performance.
What did I miss?
You will note that all the pros are mostly theoretical and all the cons are engineering challenges. This makes me suspect that possibly all of the statements are accurate to some extent.