Detecting the real cause could be quite hard. To do this you could carefully monitor the various log-files from QlikView + Windows - directly or with the help of various governance-applications respectively with own QlikView applications. If nothing could be found respectively isn't specific enough you may try to trace the live-behaviour with tools like WireShark and/or the ProcessMonitor.
If you or any colleagues have these tools already in use and are experienced with them it shouldn't be very difficult. If not I suggest to consider the above suggestion rather as a worst case measurement and think you should rather try at first any kind of exclusion approach.
For this you could duplicate the application with a new name which would remove all meta data/files and then creating a new task without any distribution. Also removing all respectively some of the access rights maybe by moving the application to another folder could exclude access-issues.
Another step might be to recreate the application which means a complete new application with only a copy & paste of the script - nothing else. This would exclude that any UI stuff like macros/actions/variables and so on have an impact and further that the application is corrupt in some way or that there are some kind of incompatibilities between the release which has created this app and the current server-release. In this regard take also a look that the releases of the desktop client and the server-services are identically.
