Information here may be innacurate, because the real notes were in my 
snesorama blog and were lost. 

- split_channels.exe -> Get an .DCS audio file (custom format of this 
game) and make other two .DCS files, one for each channel. It's 
necessary because ffmpeg tricks for decoding only work for mono files. 

- join_channels.exe -> Inverse process of the tool above. 

- Audio Specifications.txt -> Some notes for my personal use about the 
custom audio format of this game. 

- raw_header_to_correct_one.au3 -> This one will look for the generated 
mono files on the directory, create decodeable .wav files and create a 
list (raw_header_to_correct_one.txt) for further restoration. 

- correct_header.au3 -> This file will use the generated list and the 
re-encoded files to transform header back to its original form, so game 
can decode the audio. 

- split_channels_all.au3 -> Helper script to generate a .bat file to 
split channels of all .DCS found in the directory. 

- join_channels_all.au3 -> Same thing, but to join channels. 

- decode.au3 -> Generate a .bat file to decode all the audio fixed with 
"raw_header_to_correct_one.au3" using ffmpeg (included somewhere else in 
this pack). 

- encode_all.au3 -> Encode the resampled files using wavecon (included 
somewhere else in this pack). 

So I think the flow to downsample the files is this: 

Split stereo channels -> Fix header -> Decode -> Resample -> Encode 
files back -> Fix header back -> Join channels.