Collation of csv results is fairly straight forward. It uses an old dev trick that allows a program to write to files rather than a DB but avoid file locking issues (this is done by writing to individual files and having a cron job to cat them into one file every so often).
In the case of this JMeter solution, results collation is done at the end. The csv files are downloaded from the injectors, catenated into one file, sorted (on timestamp, the first field in the file) and some lines are removed. I have just used the original (excellent) solution to do this.
What I have added here is collation of more results files from the specific listeners added around the script (see the JM Results page)
collation code for csv files: