A field guide to two PySpark performance problems that hide behind healthy-looking averages. Covers what Databricks now fixes for you before you reach for salting.