Even if the crew is only in one section of the ship when it separates, the systems in the other sections need maintenance, so the crew in the third section would be much larger than for a standard ship of its size. (That is, the size of the crewed section alone.)
And because it is devoting more space to quarters, less of it's space can be given to weapons, shields, and warp cores.
Still, the other two parts, with
no crew quarters, have extra space for weapons and engines and stuff. Perhaps it all balances out.
N1N, you pointed out that the Prometheus is roughly the same size as Voyager, which had a crew of roughly 150. Fine, I'll use those numbers:
We would then be comparing the utility and cost of one ship with a crew of 150 that can split into 3 parts, or three ships each with a crew of 50.
As I noted above, the Sabre Class supposedly has a crew of 40, so Starfleet does make ships that small.
So what benefit is derived from the Prometheus's ability to assemble its three parts into one USS Voltron?
I can, actually, think of one. The warp drive cannot be run at maximum output for long periods, and in fact cannot be run at "cruising speed" for very long periods: despite the fact that the Galaxy Class is supposed to "cruise" at around 1600 lightyears/year, the DS9 Technical Manual says that no Federation Starship has ever covered 1000 lightyears in a single year, suggesting that even at cruising speed the drive is shut down for maintenance about 1/3 of the time.
It has been suggested that the reason for 4-nacelled designs like the Constellation Class is that one pair can be used while maintenance is done on the other pair. With multiple warp cores, one core could be shut down while the other is used.
Thus it is possible that the USS Prometheus can travel great distances faster than any other Federation ship while joined, because it can alternate which warp core and nacelles it is using and do maintenance on one set while still underway.
HOWEVER, this would require that each of the warp cores so used be able to put out enough power to move the entire, joined, ship at the full speed. Basically, it would need at least two (I'm assuming the fold-away nacelles on the section with the bridge can only be used while separated) warp cores as big as Voyager's, and enough engineers to be operating one while working on the other.
This version of the USS Prometheus
would need a much larger crew than a comparable ship (like Voyager), because it would need double the Engineering crew.
To a certain extent, I can see the idea that if I'm already giving it two warp cores, two sets of nacelles, and two crews worth of engineers, I could have it able to split into two equally viable ships (instead of one combat god and one sitting duck like the separated Galaxy Class).
But that leaves the question of the third section. Lifeboat for the crew?
I don't really know what is complicated about separating a ship and re-integrating again. Starfleet has been in the business of doing this for quite some time now.
Yes, they've been doing it for a long time. And it is still more expensive than not doing it. Moving parts break, moving parts need maintenance, and moving parts need to mesh correctly: a small mis-alignment of two parts is probably okay, but a small mis-alignment of four parts adds up to a major mis-alignment between the first and the last, which can lead to catastrophic failure. More moving parts = easier to break.
For this reason, you never build something with more moving parts than you need to. (Well, unless your goal is consumerism: products designed to break and need replacing. Starfleet seems to have evolved beyond that.)
So when the engineer shows you the really toyettic design he's got for a family sedan that can split into four unicycles, you ask him "what advantage does that give that is worth the extra cost in design, manufacture, and maintenance?"
USS Prometheus was designed to appeal to the part of me that looks at a fighterplane that can transform into a humanoid robot and goes "
Coooool!". And it does. But another part of me realizes that in the real world such things are rarely worth the expense they represent.