PySpark interpreter - useful debugging traceback information is lost for any error from Spark

When there is an error from Spark, the original error is not returned as output in the cell, instead a generic Py4JError is shown.

(<class 'py4j.protocol.Py4JJavaError'>, Py4JJavaError(u'An error occurred while calling o45.collect.\n', JavaObject id=o46), <traceback object at 0x7fb737ea72d8>)

While it is possible to look at zeppelin-interpreter-spark-root-node.log, it might not be accessible for a multi-user environment as it will require remote access to the host running Zeppelin.