I'm sorry, but I'm not proficient at AviSynth and I wanted to use simple tools to do it. I have the best luck with TMPGEnc. You set the size (720x480), 16x9, encode mode to "non-interlace", frame rate to 23.976 fps (have to do it in that order). Under Advanced tab, checkmark " Do not frame rate convert ". GOP structure should be Max 18. Shouldn't need to make any other changes. When finished, process it through DGpulldown 23.976 --> 29.97 to set the pulldown flags.

I then used Womble MPEG video wizard as a check because it will tell me to 4 decimal places how long it is in seconds (might need to check in minutes to get that part, then seconds, because minutes gives the decimal as frames). I then used GoldWave's time stretch feature on the wav and set the time stretch feature to match the time (setting warns that it will change the pitch, which you want). You can then convert it to AC3 using whatever tool you have.

When everything is complete, i checked it all out in Womble to make sure the lip sync was good, then MUXed it with whatever authoring program I was using.

For some reason, my copy of Empire was corrupted and had a slightly shorter soundtrack than video track, so I had a lot of problems getting it to exactly the right length. Jedi didn't and worked first time through, as has every other video I've processed.

Also. I seem to recall having to encode BE and RTJ in two steps. First was all the settings I mentioned (but a much higher bit rate for first encode) except frame rate (kept as 25fps) and I also deinterlaced it in TMPEnc advanced tab; then processing it again, but with the frame rate change. For some reason, doing the deinterlace and frame rate conversion at the same time was processing each frame twice (that is, double printing). I don't know if it's because I've only worked with progressive material since then, but I've never had to double encode anything since.

You can adapt to whatever program you use, but be sure when you change the frame rate to only change the frame rate number (that is, 25 to 23.976) and that it does NOT "frame rate convert". It seems counterintuitive, but that setting tries to alter the frame rate but make sure the overall time length is the same. You don't want that. You want it to end up longer and the sound out of sync because you are going to stretch the sound out to match the video.

I also used Adobe Audition on the WAV (Gold wave might do this, too) and selectively re-pitched sections of the commentary voices. Using the spectrum analyzer display makes it easier.

Let me know if you need any help.