In highly concurrent situations, multiple clients may request the same content from an NXRM proxy repository. In these cases, the parallel requests significant churn on the Asset record, with unnecessarily large numbers of blobs created as a result of transaction retries fighting over which blob to attach to it. This could occur in multiple HA nodes using the same blobstore, or on a single node (multiple threads). Because the soft-deleted blobs from the unsuccessful tries linger until the blob store is compacted, this can also waste storage space.
Detect when a transaction is attempting to attach a blob to an Asset when an identical blob already exists on that Asset. In those cases, ignore the attach request, and retain the existing blob.