It's notable that projects like spark could have used Jython for its JVM/python interop, but chose py4j instead, consigning pyspark to some rather abysmal performance penalties under many conditions.