Whichever method you choose, you will probably want to start by designating one track as the master reference. Usually, that will be a drum track.
When I've done this in the past I've used the kick drum, which, even if it's not on its own track is usually the most reliable reference. If all you have is a single drum track, you can clone it and filter out all but the lowest frequencies. That will allow you to generate a corresponding MIDI track that follows the kick drum. There will be some spurious hits in there but they can be manually edited out.
If for some reason that method isn't possible (e.g. there is no drum track) then you can record a MIDI track by manually playing along with the recording. The idea is to end up with a track that SONAR can easily line up its tempo map to.
Once you have a timeline that matches your reference track, you can now easily jump to a particular beat and see which tracks are ahead of it, behind it, or close enough to it. Bear in mind that not every beat that's not exactly on the mark needs correction - remembering that will save you a lot of grief because every edit has the potential to make things worse, not better.
Once I have a guide, the next step is to determine where the most noticeable timing errors are and whether cut-and-nudge or AudioSnap is the best way to correct each of them. I'll usually cut the tracks up into sections, e.g. intro, verses, choruses and significant cues. It's important to cut each track individually. For sparse tracks, look for silences for a place to cut. Even dense tracks will have quieter moments where a cut can be made with minimal interruption that might be noticed. Be sure to use short fades on each side of the cut, which will eliminate the need to find a zero-crossing (still a good idea though).
Next, I classify clips into one of three categories: good-enough-don't-touch, slightly-off-minor-tweaks, and what-were-they-thinking. For the first category, I fight the urge to make everything perfect and leave them alone. For the second category, AudioSnap works well for anything with identifiable transients (drums, percussion, bass, vocals) and I prefer to try that method first. If AudioSnap can't isolate transients, or if the timing errors are so far off that A/S can't render edits transparently, then it's cut-and-nudge.
There have been cases where all the above just wasn't practical, e.g. the drummer was apparently playing a different song than the rest of the band. If that happens, ask yourself just how much effort you're willing to put into the project. No matter what, it's always do-able, but you'd better be being paid by the hour.