There are at least two FOR loops that take a lot of time to complete, probably because of the large number of files the script is processing. My first suggestion would be to move each loop to a separate document, and run both in parallel. With a publisher license, tasks could be programmed quite nicely to run in parallel and continue with the next data stage when those two documents have both finished reloading. Without Publisher, you may have to figure out which one is taking more time at all times, and let that one initiate the next step.
Other suggestions:
do all processing in memory. Doing various table manipulations while reading from text files, pulls down your reload speed. You may need to check on the amount of RAM in your QV server..
try to categorize and throw together all file data into few QVD's. QVDs load many orders of a magnitude faster than an xml or csv file. Again you can use separate QVW documents to consolidate all text files before you start doing filtering and aggregation in your current QVW.
Check the TaskLog. The TaskLog information can be found in two places:
In QMC->Status->Tasks, select the task you will be investigating, click "Show Task Details" (top right corner), select the failed reload time in the listbox on tab "Log" and check the end of the log details.
On your server, open Windows Explorer and navigate to C:\ProgramData\QlikTech\DistributionService\1\Log\. Select the directory with the correct date and open the subdirectory with the timestamp and name of the failed task. There should be a file called TaskLog.txt. Open it.
If you want further feedback, post the contents of that TaskLog in this thread.
