The current implementation used to collect metrics related to the frame time could be improved reducing the number of allocations used when posting a new callback to the choreographer.
This will reduce the SDK memory footprint and will improve the performance because the GC will ve invoked less.