The changes, possibly for WT-671, or other LSM related changes earlier this week have resulted in a consistent 20% hit on the perf-populate-lsm test (the load time portion). I changed the plot to show the last 100 data points and it clearly shows it. Take a look at the plots in Jenkins. (And that test is one that really is very consistent).

Is this penalty expected with some of the new locking or other changes recently?