The 130k overflow performance test is showing a regression of about 30-35% in its read and update throughput. I have attached a screen shot of the plots and the changes.

The apparent "recovery" of the performance in the chart is due to running on a different branch that may not be caught up with develop.