In order to confirm the currently vague and more intuition than hard facts based suspicion that there could something wrong with memory consumption of the JNI LevelDB embedded in ODL, skitt and I together had the idea that perhaps we could get jluhrsen to run CSIT on the pure Java instead of native JNI LevelDB, for a test..

This COULD have other problems (I'm not clear / doubt how much testing the alternative pure Java LevelDB implementation has gone through), and the goal here isn't the get fully passing CSIT - just to see if, miraculously, the ODL JVM kill does not appear anymore just if we switch this...