Stream Analytics compatibility 1.2 - PARTITION BY PartitionId

Will the PARTITION BY PartitionId statement still ensure parallel processing (and make sure that input partition = output partition) for compatibility level 1.2 when we don't specify a partition key column?

Our situation:

• we have 2 input eventhubs (both 16 partitions) and an output eventhub (also 16 partitions)
• we have a stream analytics job that joins messages from these 2 input eventhubs and writes them to 1 output eventhub - these messages are also joined on partititonID
• we use compatibility level 1.2
• we have NOT specified a partition key column
• we DO however still use the PARTITION BY PartitionId statement in our query

Question:

According to the documentation, I would expect the messages to be processed embarrassingly parallel. Where I assume that the messages will end up on the same partition on the output eventhub as their originating partition (i.e. messages from input partition 1 will end up on output partition 1).

However, is it a problem that we use compatibility level 1.2? In the documentation I read that PARTITION BY PartitionId is required for compatibility <1.2, however, does this still work for 1.2 or do we then need to set an partition key column? (we don't do that yet since our partition key is a combination of multiple columns so setting the output partittionkey to be the same as the input partition key was an easier solution for us)

Our simplified query:

SELECT
	evh1.timestamp,
	evh1.some_column,
	evh2.timestamp,
    evh2.some_other_column
FROM
    [evh1] evh1 TIMESTAMP BY evh1.timestamp PARTITION BY PartitionId
    JOIN [evh2] evh2 TIMESTAMP BY evh2.timestamp PARTITION BY PartitionId
ON
	<some joining logic>
	AND ppv.PartitionId = batchid.PartitionId