There were about 1% failure on the Drools S3Ps, the failures seem to be related to DB access. The DB is being use heavily, there is a very short pause between executions. This may not be a realistic scenario in real deployments. However, this is worse than in previous releases. This has been observed in going from Jakarta to Kohn. Perhaps we could use the lab to see how the testing could run with a drools/apex policy scenario running.