What is the suggested pattern for representing a table which can be updated by multiple jobs? For example, a table which tracks metadata of files downloaded by jobs. We have been representing these as separate SDAs (i.e

job1_download_metadata

,

job2_download_metadata

, etc.) derived from the same underlying op (via

AssetsDefinition.from_op()

), but would ideally like to unify this into a single

download_metadata

SDA.

One way to model it could be using asset partitions, where the “download_metadata” can be one asset and asset materialization run by different run can be in different partitions.