In order to make this benchmark viable again for regression measurement, the cause of the modality starting in 10-b29 needs to be determined and fixed.

Java should have consistent, predictable performance on Linux-x64, so fixing the modality is important for that as well.

Once the cause is determined, this bug should be re-assigned to the appropriate team.

b29 had no hotspot changes, the the cause of the modality is in the JDK.