I suspect I'm not getting the usefulness of the feature but I'm stuck in that old-school "sound-on-sound" world (like when you punched in with tape).  Not hearing what's been recorded already for punching-in is awkward- at least to me.  So I've been switching all my projects back to Sound-on-Sound.  I watched a uchube video on it and I can't say it wow'd me to want to use it.  Anyone totally turned on by this feature?  What am I missing? 

this will help you significantly improve your comping skills
 
http://www.cakewalk.com/CakeTV/SONARU.aspx/Get-Started 

It's so fast and seamless! When I have a vocalist recording, I let them sing through the song several times. Then I use the techniques shown In the videos to isolate and combine the best takes in each phrase. I am not exaggerating when I say that clients have been astonished at how quickly I have combined takes to create a finished product. It is most useful when dealing with multiple complete takes. I think it is a well implemented improvement, but I don't expect that it fits everyone's work flow.

If you do a bunch of complete vocal takes, as James says, the new comping thing is very handy and fast. I watched the video (check chuckebaby's link above - I'm not allowed to post links), and they were bragging about it so much that I decided to try it. The technique that makes sense to me is the one described in the video beginning at 2:19, where you go through the song (after you've sung all the takes) and split the whole thing up into phrases. You can then audition each phrase and quickly promote the best take to the top (comp) track. It works beautifully, if you're willing to record this way.
In the past, I'd do a few takes, pick the best one, then punch in the parts that needed fixing. Probably a lot of singers will want to do it this way. If you work with singers other than yourself you know you'll have to do it whatever way they want to do it, but I can tell you the new comping tool works great.

I recently did some takes of me playing a trigger finger for a hi hat part that was four measures long. I did about five takes. Then did some other work, came back and recorded five more takes again. Went to the store and came back and did five more takes.

Using the comp tool, I was able to get enough variations of hi hat patterns for the while song in a fraction of the time it use to take me.

I'm sold on it. But it did take a little effort on my part to learn the methodology.

Sacalait
Not hearing what's been recorded already for punching-in is awkward- at least to me.  So I've been switching all my projects back to Sound-on-Sound. 

this shouldn't be happening.
in over write mode you should be able to clearly hear the track before punching in.
matter of fact in any recording mode.
you must be missing something in your settings, more info might help your situ.

Chuckebaby!  This is how I've been doing things:  I record, let's say, a vocal track.  We listen and the artist says 'I want to redo a particular phrase.'  I usually split the clip at the spot of the phrase and slip edit that portion out so the artist wont hear it when he's overdubbing.  I'll give them some preroll so they know where they are in the song- like a phrase before the punch point.  I instruct them to sing-along because it will be a more natural sounding punch once we hit the punch-in spot.  think of using a tape machine....  yeah, I'm old-school... but it the same process.   On all previous versions, it worked fine for me- and STILL does!  If I switch to "sound-on-sound."  Maybe I'm not using the feature as it's designed to be used- like some of you say- cut a bunch of takes WITHOUT listening to the previous.  THEN go through the various takes.  It's not exactly how I do things but I'm open to making my job more efficient. 

If you enable Dim Solo in comping mode, you can hear the other tracks.

Thanks for the suggestions.  I'll look further into it.

I also have to switch to sound on sound if the singer wants to hear him/herself the sentence before the punch. If it's on comping mode, it mutes the previous takes and creates a new lane. I think that's intended behavior but if it's somehow selectable that'd be great. I really should read the manual sometime.