The symptom is that tasks take longer and longer tor run. A restart of the Qlikview Distribution Service will temporary solve the problem.
This has to do with the number of something called "Pinned objects". This number is usually in the region 0 to maybe a couple of hundred. At some customers it has been seen growing to millions. The number keeps growing until the Distribution service is restarted. As the number increased, so does the task execution time - until QDS is restarted. This can also be seen as the Distribution service consume a lot of CPU even when no tasks are running.
To check this symptom view the number of pinned objects.
This by open performance monitor, right click to add counter. Expand .NET CLR Memory and select # of Pinned Objects- Also specify the QVDistributionService. Click OK.
The Processor time counter is automatically added and has to be removed by right clicking again and selecting properties. Then remove it.
Also in the properties adjust the scale. For example if the maximum number of objects is around 600.000, set the scale maximum to 1.000.000.
The above picture shows the maxium is only 156 which is fine. But if the number is growing into hundred thousand then it is probably this issue.
This issue is caused by a memory leak in Microsoft FileSystemWatcher in .NET Framework 4.
Install Windows update 2600217 (http://support.microsoft.com/kb/2600217). This should be installed with standard Windows Update as part of "Microsoft .NET Framework 4 Client Profile".