Wouldn't just having multiple staging tables or databases across multiple physicals be good enough to parallel load the data and process it before merging into one primary table? You're still going from say 4 lanes to 1 lane, but if the data is processed, refined or summarized in the 4 lanes before merging to 1 lane, there is still augmentation of the workload across those 4 tables that can save you time before hitting the 1 physical table regardless of table partitioning?