I use AudioSnap for only a few things: lining up tight vocal harmonies, correcting bass guitar sloppiness, lining up hand percussion, and for generating a tempo map from a live drum track. The latter works well if you have a separate kick or snare mike to key off of. It's the only scenario in which AudioSnap is a simple "click here for magic" operation.
Bass works best if you only use initial transients to quantize a note start to the beat. I'll typically delete all but the first detected transient, so I'm essentially just moving whole notes around with little or no stretching. Sometimes it's easier to just split-and-nudge instead of bringing out AudioSnap, if there are only a couple notes that need tweaking.
The vocal alignment procedure is labor-intensive because you have fewer reliable transients to lock onto. I typically delete 50-80% of the detected transients and often move the remaining ones visually or insert my own. This works OK because for vocals you're usually only trying to line up consonants, starts of phrases and durations of held notes. It's rarely worth the effort for a lead vocal alone, but it pays off for multi-part harmonies.
At least in these limited applications, AudioSnap is a wonderful convenience that really works. I just wouldn't trust it on anything that didn't have well-defined transients (e.g. strummed guitar).