I've got a bass guitar track showing a peak of +0.4 dB at a single spot in the song.  Everything else in the track is good, peaking at about -3dB in a few places.  I've isolated it to a single note, but the waveform in track view doesn't show anything that looks like a peak, even with time scale zoomed in all the way.  On this particular note with the peak, the scale on the waveform shows a max of about -9dB.  I've got Concrete Limiter on the track already, I assume I could crank it down more and get rid of the peak but I don't want to change the entire track for this one peak and then work to get the sound back everywhere else; but mostly I don't understand why it shows a peak.  It sounds fine, no audible unwanted distortion to my ear.  
(A) why does it show a peak but the waveform looks nothing like a peak?  and
(B) do I really need to worry about getting the peak down on this single track?  (My guitar bus is - 6.6dB at this point, with the bass and 3 other guitars.) 

Forgetting all about the "why" for a moment, if it's only peaking (and not clipping) at track level and not at the master, then you have nothing to worry about in a floating point system.
The only time sound will be degraded is if you clip your converters.

OK, thanks.
 
Now the "why"...?  (Curiousity)

So, you've got some fx on the track, right? EQ, Compression maybe? You say a limiter. I have found sometimes that I'll clip a frequency. Most of the moment is well below the clipping point, but maybe at one frequency I go over the top. Is that possible? So maybe a little fret rattle from a firm attack which the compressor lets through because of the attack setting?
 
I'm just WAGing here.

Another reason why you need a separate editor such as Sound Forge or Adobe Audition etc. I like opening up tracks before the mixing stage in an editor. There is so much that can be accomplished there. Firstly they do these things better than any DAW for some reason. Stray peaks are easy to control. I tend to edit dynamics mainly in the editor and not touch EQ so much. I think that is best left alone so you can EQ things in context with others.
 
A bass track might be sitting at around say -16 dB rms for most of the track but there might be 8 or so peaks coming close to or even hitting 0dB FS. So you tame just them down with the editor, then you can lift the level of the whole track up to say -14 dB rms or higher. (you do not subtract gain either for peaks, you limit the upper level of the peak to match the rest of the track)
 
You can clean up noisy areas in unwanted sections. (eliminating gates now) With vocal tracks it is easy to edit the quieter phrases up a little and the louder ones down a little to even things out much more.
 
It is important with editing to not change the length of any audio track for obvious reasons. Once you have edited anything you resave the track as is but add the extension ED to the title. That way you know you have modified the track in some way. Then you remove the existing track in your session and drag the ED version in instead.
 
The good thing is that once you do some editing work then you often don't need any (or very few) plugins on that track now in order to make the track more normal and sit nicely in the mix. You have achieved it all using the editor instead of automation and plugins. Your CPU will love you for it and the track will still sound better as a result too. How many plugins do you have on a track? Ask yourself WHY and is there a way to achieve a good result without them. Editors can really help here. There is also a level of precision with an editor that no plug-in can achieve too.

gswtiz, it looks like you're correct.  I turned off each module in ProChannel one at a time, turning off the gate dropped it to 0dB; but turning off Concrete Limiter with the gate still on dropped it to -1.8dB.  I suspect you're right that it's the attack in this location.  I'm not changing anything, but it makes me feel better to understand it.
 
Jeff, your approach is very intriguing, but I'll admit it's way over my head.  I'm more on the artist side of things.  Still appreciate it though -- always nice to learn how much more that can be accomplished.

I have also reduced the odd stray peak with Process>Apply Effect>Gain or also automation

M_Glenn_M
.... or also automation

+2.98768574768696857474636769

bapu

M_Glenn_M
.... or also automation

+2.98768574768696857474636769

Nah, it's more like +2.987685747686968574746367691

MorganT > Jeff is talking about using other Apps to make changes to the waves off-line and then to mix in Sonar.
 
Advantages include you getting processed (kinda like normalized but more specific -- in other words, the tracks might have a standard RMS level but not be using every available bit at the loudest moment). You can apply specific EQ and compression on specific wave cycles or transients. You can change the amplitude of certain waves by reducing the amount of different portions of the wave to make the adjustments as transparent as possible and as true to the original as can be achieved.
 
You can then apply all the FX and make all the adjustments and render a new wave file. The amount of processing required is some what irrelevant since it is the only work being done at the time. Then you render the new wave form.
 
Once you have rendered new wave forms for each track with a similar RMS level you can mix the remaining tracks more easily. Because all the tracks are 'gain-staged' or 'processed' or 'normalized' (all these have more specific meanings but I'm using them here anyway). In general, you know have minimally compressed tracks carefully adjusted to each have the same RMS levels.
 
http://www.gearslutz.com/board/mastering-forum/173236-what-difference-between-rms-peak.html
RMS is effectively an average of peaks over time.
 
Some people swear by staging RMS levels across various tracks before mixing the tracks.
http://www.kvraudio.com/forum/viewtopic.php?t=135884
 
In a lot of ways, this pre-processing is an effort to make the mixing easier. It might really help. Obviously, Jeff and others are sold. I have spent time watching RMS Levels on tracks as I mix. It's interesting info to work with and it does seem to help me get to a better mix faster, but I don't do off-line rendering in order to get the job done myself. I just send the tracks to RMS meters pre and post to get a view of what I'm doing to them all. It works pretty well if you don't have too many tracks.
 
One of the risks of pre-processing is over-processing the tracks. In my opinion, an example of over-processing a track is making a final mix at one level of compression and then deciding to make the whole track quieter for the CD. Worst would be to re-expand the compressed track!! why not double back and bounce the track with less compression to get a different RMS level on the final mix??
 
This is probably way more info than you wanted, but they are my thoughts regarding Jeff's comments (which are great comments btw). Jeff is pointing you to a really super great lesson in mixing. He is telling you how he does it which is with other software outside of Sonar. It is not a requirement to use something outside of Sonar to use the mixing technique. So, check out his lesson. Play around with RMS on the various tracks you are mixing. See if it doesn't help you get a better mix faster.