Yeah, sorry, I could have been clearer here.
I'd suggest that each fixed camera has to do three things (in order of importance, with 1 being most important):
1) Provide readable environments with clear paths of traversal and affordance opportunities for players. This, I think, can be proven by using rudimentary geometry as seen in the Parasite Eve screenshots posted above. Like greyboxing, your level should be fun to navigate and play at this stage.
2) Convey information to the player through established cinematic language (eg. elevated angles emphasise the weakness of the Player Character, refusing to move the camera with the player shows their lack of power etc). Camera angles need to change with the >30degrees rule, though they will often ignore the 180degree rule.
3) Aethetics - framing, prop design, lighting, camera lens, post processing, particles, colour palette, etc. Some elements here (palette, for example) might be worth figuring out before building the environment (1), but for the most part polishing stuff like this could be left towards the end of the fixed camera design.
... I'll be honest, I'm still thinking this stuff out. I might swap some bits around! But each fixed camera should be humming with intentionality, and so I tend to think of 1 and 2 first. ...Just thinking about process, one might design the camera positioning with 2 in mind at first, refining with 1 then polishing with 3. =/
I figure this'll get mixed up a bit during production!
The swivel camera - it can be nice to maybe imply that somebody's watching the player, but... A key reason to go with fixed cameras in the first place is that you believe an intentional, authored choice of camera angle can impart information about mood, power dynamics, relative knowledge, emotion, irony, goal progression, attitude, stability and more to the player without going right out and just stating it. You tap into the Kuleshov effect by changing the camera when the player moves into a different zone - they *will* infer meaning as they encounter this.
But as soon as you put the camera on a swivel and have it auto-track the PC without vignetting, all that goes out the window. The camera is just a boring slave of the player - the only meaning it can impart is that which the player gives it by moving closer or further from the pivot. It says the developer wants to achieve the tone and atmosphere of an Alone in the Dark 1 or Resi 2 and has technology beyond the dreams of the makers of those games... but has absolutely no understanding of the craft or conventions they're failing to ape.
...
Blah blah blah rant. =D As I say, I've yet to test a lot of my theories. But it's interesting, no? Thanks for taking the time to think and read, Ohmygiggles!