I vary between 1-4 tracks playing at all times normally, but from time to time I'll need more if there are overlapping parts or if something specific should creatively call for more layers. I generally set up my tracks like this (and more if necessary, following the same organizational scheme):
Vox C1
Vox C2
Vox L1
Vox R1
I also normally have the vocalist do one pass straight through the song first (especially if I don't already know the song), that way I have some reference as to how it's laid out and how to approach it better, what to ask the vocalist to do, etc. Work back through the song from the beginning, redo most (if not all) of the initial pass by section/phrase, add layers, throw in my ideas (accent powerful/important lyrical passages, etc)...and basically by the end of tracking, I have all the layering and panning all set based on which tracks I recorded to. It just seems wrong to me to not have the overall production approach all decided on by the time tracking is done, I think it's gotta be all worked out while the vocalist is there (if not before) so you can make sure the song turns out as good as possible. Also then this way I don't have to do much (if any) panning automation, everything is just already placed where it belongs, and it's mostly just volume and FX send automation to deal with. I'm pretty much thinking about screams/growls as I write all this, but the same rules apply to singing!
ESPECIALLY with screaming/growling, I think it's really important to vary between a single track at times (C1), and various layers (C1+C2, or C1+L1+R1, or some pseudo-stereo L1+R1 or C1+L1+R1, etc), in order to really add movement and vibe to the songs. No matter how good the vocalist is, I think it gets really boring hearing a whole song worth of growls just plopped into the center of the spectrum without any variation and contrast across the stereo image and in the way the vocals are presented to the listener. I'm really big on having vocalists add several identical layers too...the idea isn't to always have the layers always sound like different kinds of voices, but to have the sum of all the parts amount to what sounds like a single larger-than-life voice.
Anyway, every song is unique, but that's my general approach.