From data safety and consistency perspective, this code makes a lot of sense - we make sure each backup has its own attachment holders. It's slightly worse about performance due to multiple Blob service calls, however this code won't be run frequently.
A potential improvement would be to use the batch-holder endpoints, but this is a follow-up for a separate diff, it can be deferred.