Your multicam sequence will always pull audio from whatever is "track 1".
Whenever I have audio I need to bring in from multiple cameras, I'll sync everything up by timecode or audio/visual sync and then export my audio to a third party app like Audacity, do whatever audio editing I need to do, then bring it back into Premiere as a stereo wav file and set it as track 1 (you may have to create a new audio track... make sure you specify it to show up "before track 1") Then the multicam editor will pull your audio from that file.
Make sure your audio tracks are targeted in your source sequence (this used to be an issue in CS3 and CS4, but I haven't personally noticed it in CS5).
I regularly work with 4 cameras and 4 audio sources, so I can tell you it works fine (yes, some audio tracks drop out at various points while others continue). I haven't had to export/import audio except on occassions where I have more than 4 audio tracks, or where I had other reasons to do so. More often than not, I just cut the audio inside Premiere and it works fine.
Try targeting those tracks. Also, you should be able to ALT+select your audio track in the multicam sequence and choose which audio track you want. You also have choices for audio-follows-video in the multicam monitor flyout menu.
Now I WILL say that the multicam workflow DOES have some issues with audio (rendering, losing track of the rendered files, rendering again, no display of the source waveform). But those are long-standing issues and don't prevent actually editing the audio in the way you are describing.
Hope this helps!
So are you saying that the "Track 1" thing is no longer an issue in CS5? That's really nice! I've only done one multicam edit since I got CS5 and never even bothered trying anything new since I've done it with the export/mix/import thing for so long. I'm looking forward to my next multicam, now!
Thanks for the replies.
I solved my problem by backfilling track 2 to the start of the source sequence, which then allowed me to successfully select Track 2 in the multicam sequence. This suggests that davidbeisner2010 in the first reply is not describing accurately what happens in CS5.
I would still like to see the multicam sequence include an option to take a submix from the source and will be submitting a feature request to that effect and also to suggest that audio track swapping in the multicam sequence is facilitated separately from the pictures. This would give much more flexibility in dealing with multicam edits.
The audio selection in multi-cam could use some improvements.
An easy work-around I found is to do the following: When nesting the initial sequence into the mutli-cam sequence, do so twice resulting in two video and audio track pairs. Then enable multi-cam on one pair for cutting. Delete the second video track, and mute the mutli-cam audio track. That leaves you with a copy of the audio which should be perfectly synchronized, but where you can submix and know exactly what you're getting without having force an audio track to be in a particular position.
What works for me is to CMD + double click on the clips in the sequence. This opens them up in a new, temporary sequence. I can then do cuts and edits on the audio channels to my suit my needs. You'll notice when you do this that channels outside of A1 are always muted by default. Once you've edited the audio (or video) in this temporary sequence, close it and your changes for that multicam clip are "hard coded" into the clip that's in your main sequence.