It turned out that the bottleneck of the annotations visualization data load function was the point where thousands of annotations were loaded one by one. After changing that with a one time multiple_load some the total execution time of that function reduced more than 5 times (from 4.2s to 0.8s, for 1048 annotations).
It turned out that the bottleneck of the annotations visualization data load function was the point where thousands of annotations were loaded one by one. After changing that with a one time multiple_load some the total execution time of that function reduced more than 5 times (from 4.2s to 0.8s, for 1048 annotations).