When I run a multi-partition backfill as a single job, the asset materializes for each partition - does this mean the asset code is being executed for each partition? My goal was to only run the asset op once, using the partition range, so I'm not making multiple requests to the API (which accepts a date range). My asset is using context.asset_partition_key_range_for_output(). To make things more confusing, each materialization has the records for all assets.
e.g. each partition should have 26 rows, but when I backfill for a 3-day date range, instead of getting a single materialization of 78 rows, I get 3 separate materializations of 78 rows, which I assume means the asset code is being run (and the API is being hit) three times.