The performance in the Custom Report should be very close to the performance of a single native table in Qlik Sense. The reason is that the Custom Report actually use the native objects and hypercubes of Qlik Sense. The objects are created when you select the dimensions and measures. If you feel there is a performance issue with Custom Report please check these two things:
1. Create a native Qlik table using the exact same dimensions and measures as you select in the Custom Report. Any difference in performance? (The result will be cached by Qlik engine so you have to restart between tests.)
2. In the extension properties find the setting for “defer layout update” (under Settings/Experimental Settings). That mode is made for handling larger data sets and will wait for the users to accept changes in dimension/measure selections before calculating the chart.
Of course if you still experience issues with performance that you think is extension related, send an email describing the issue to email@example.com. If possible include a sample application.
Always start the custom report with a smaller selection of dimensions and measures in the default state. This way the initial load will be shorter.
When creating the master item table use a very small selection. (Since that table typically has a lot of dimensions and measures it can be slow with larger data volumes.)