So I have this piece I'm almost done with that has been a real beast in its complexity. Time sig change every bar or so. It's only got about 20 tracks though, about 10 instruments. Now that everything is finally laid down, I'm just about to start mixing it, and suddenly I realized it would be much better about 5% slower.
 
I've already frozen or bounced some tracks to audio, and I can just re-do those, but there are a LOT of vocals, as it's a patter song with some harmonies and double tracking. How do I adjust the audio to the slower tempo? My vocal tracks were in a lot of clips, and when I simply changed length and chose to Stretch Audio, the resulting audio sounded fine, but the clips didn't simply move to where they would have been had I recorded them at that tempo, and instead jumped all over. I undid and tried bouncing each track to a single clip, and then for good measure also extended each bounced clip left to the very beginning, but still the audio, while it slowed down correctly, moved to unexpected places in the timeline.
 
How am I supposed to do this?

The new Melodyne 4 maybe?

SONAR does not have a built-in function for scaling a tempo map. Below is a workaround I've posted before. It's awkward because you have to work in SMPTE time. Also you should avoid trying to scale only part of the project; IIRC, there are issues with the way SONAR handles that.
 
1. Enable autostretch (a.k.a. Clip Follows Project) on all audio clips.
2. Select a MIDI or audio clip that runs the length of the project (or select in the timeline).  
3. Go to Process > Fit to Time.  
4. Convert the current Thru time to total frames.  
5. Divide that by the desired factor (e.g. 1.10 to get a 10% increase, 0.90 for a 10% decease).  
6. Round and convert that new value to minutes:seconds:frames.  
7. Enter that as the new Thru, and select Modify by Changing: Tempo Map.  
8. Click OK. 
 
 
 
 

OK, some new information. I've solved the problem, but I'm not sure why I had to do it this way, and this seems like a good candidate for an authoritative answer on the best practice here.
 
By Selecting Process - Length... I was able to adjust the tempo of my audio tracks. (Oddly, the stretch audio selection of Radius Solo fared better than Radius Solo (Vocal), even though these were vocal tracks. Whatever.) Anyway, if I adjusted the tempo of each clip separately, then everything was fine. If I tried to save time by selecting multiple clips and adjusting the length all at once, trouble ensued. Also, if I bounced clips to a single clip and then tried adjusting tempo, things also went haywire in terms of timing.
 
My foggy understanding of this is that since I had adjusted the project tempo before changing the clip length, somehow that messed things up, because adjusting the tempo moves clips to the same measure-based starting point in the timeline. So, though it is done now, what I should have done as fas as I can tell is bounce the audio clips first, and only then change the tempo, and lastly alter the (now bounced) clip length. Right?

jkoseattle
My foggy understanding of this is that since I had adjusted the project tempo before changing the clip length, somehow that messed things up, because adjusting the tempo moves clips to the same measure-based starting point in the timeline.

Yes, the default 'Timebase' for both audio and MIDI clips is 'Musical (i.e. M:B:T) so clip start times follow the M:B:T timeline. Normally this is what you would want if the audio is already aligned with the timeline, and Autostretch was enabled on audio clips so their length follows tempo changes as they're made.
 
But if you changed tempo and then did the length change, you would run into problems with clips that crossed tempo changes.
 
The method I outlined avoids all of that. You might want to go back to the original version, and give that a shot. Depending on the clip relationships in the project, I think it would  be virtually impossible to correct the problems created by changing tempos and manually stretching in two separate operations.
 
Using Autostretch is very straightforward for changing the tempo of a project with a fixed tempo. The Fit to Time procedure is only necessary in order to get multiple tempos to scale proportionally.

OK, I followed your instructions, fortunately I understood exactly what you were trying to do. Still didn't work after all my calculations. Even with single bounced clips and all, certain things weren't lining up correctly. And for some reason in my original project with the wrong tempo, a wonderful and repeatable audio artifact has cropped up. Right in the middle of my piece I now get this huge loud epic twang of doom (TWAAAAANNNNNNNNNNNNGGGGG!!!!!) with a lovely long decay on it. Stopping the audio engine stops it. It's actually a pretty terrific sound and repeatable. It's rare when a Sonar glitch makes me laugh out loud rather than grunt in disgust. (I had to record it on my handheld recorder), but alas, a huge epic twang is not appropriate for my cartoonish patter song :-). I give up, I'll re-record. Thanks for the help!

This was a common issue back in the days of tape, where the artist wanted to slow down or speed up a song. Many (most?) early multitrack recorders did not have variable speed. The solution was to mix the song down, then use a two-track tape recorder with variable speed, and record playback from that deck to another two-track.
 
What I do with SONAR is mix, then if it doesn't matter whether the pitch is a little bit sharp or flat, bring the file into the loop construction window and do the variable speed trick I've mentioned before that does NOT involve looping and adds no artifacts. Several of the songs on my new Neo- album have slight speed variations done in this way. If pitch does matter, I shorten or lengthen the file with DSP. 
 
There are two big advantages of this approach.
 
1. You don't have to apply stretching to every file used in the mix.
2. If you change your mind, you can make changes while still leaving your project completely undisturbed.

jkoseattleIt's rare when a Sonar glitch makes me laugh out loud rather than grunt in disgust.

 
Your description made me laugh out loud. I think I've heard that sound myself. Sorry it didn't work out for you, but thanks for the laugh. 
 

If late in the project, Craig's route is what I would choose as well. A pre-master mixdown can be modified at will in a new project without the headache of too many irons in the fire. Plus can have tempo changes without alignment concerns.

Downside to this is that mix changes require this to be redone, but if this is a simple "whole project" shift is not only simple but allows you to try different tempos easily.