For me, there's two different types of reverb approaches.
1. To supply a blatant reverb effect that is not meant to emulate a real room or space. Think dub (REAL dub, not dub step) or when you're producing an eighties track. There are aesthetics to the sound that are part of the production style. In dub you'll find that the drums are usually very tight and dry and once they pull you in with that immediacy, they'll trip you out with a single reverb stab on the snare to bring you back out to space. Same goes for degenerated and filtered tape delays. In these approaches, algorithmic or mechanical (plate, spring) reverbs seem to work best.
2. To supply spacial cues that are otherwise absent in the original. For me the primary use of this approach is to give depth to the mix and play with the ear/brain's positional cues. In other words, to emulate real life. Convolution or impulse reverbs work best for me here. If I want to emulate room mics or give short ambiance to the vocals, this the approach I take.
In terms of what I usually use, I'll have a long and a short reverb for vocals. I don't use a "snareverb" or any other kind of reverb on individual drums (unless it's a blatant effect or for programmed drums where there is no bleed) because I create virtual room mic tracks by running the entire kit through a room impulse - balanced adequately - and then printing the result to a split stereo pair. 9 times out of 10 it's one of the Bricasti impulses by Samplicity or Acousticas. I then compress and mangle them just like I would do to real room mics and then I send them through the drum bus to be compressed with the rest of the drums to glue it all together. I hate the sound of compressed drums with pre-sent drum reverb mixed in on top. It just sounds like it's blanketed over it. Alternatively, I'll use VOS's Epicverb, which is algorithmic, but sounds great.
For long vocal reverbs I'll use a Plate 140 or some kind of EMT impulse along with a very short ambiance, also usually a Bricasti. I use the ambiance to give the vocal some space and initial reflection and then mix the long vocal in to taste, depending on the song. I've never been one to let reverbs stick out like sore thumbs but rather take the approach of giving the IMPRESSION of space without someone noticing the reverb outright. More often, a little goes a long way.
The rest of the time I apply reverb on a case by case basis. Sometimes a guitar needs it, sometimes it doesn't. If an acoustic guitar sounds too dry because it's on it's own, I'll add some, but in that case it's the same reverb I use on the voice. Acoustic guitars like long plates or halls when they're on their own, imo.
I don't think I have ever used reverb on bass. Maybe if I ever cover the Seinfeld theme...
Keys...well...they often have reverb built into the patch. A good concert hall reverb on a piano couldn't hurt. Or a jazz club.
I also try to get away with as few reverb channels as possible. Too much reverb can wash a mix out. Not a fan. I would way rather capture a great sound at tracking and be very judicious with applying reverb. At the end of the day it's about the production's vision and what it needs. Having a producer helps. He can guide you towards making the right reverb decisions,...if he knows what he's doing. Knowing how to play with the three dimensions is also key and why using EQ to control the vertical, panorama to control the horizontal and reverb to control depth of field is so important in mixing. Our ear is built the way it's built for reason and if we use reverb correctly, we can fool it quite effectively.
That is all I have said or have ever wanted to say on the topic of reverb.
Cheers
