There are cases where the content/length (size of the main blob) is not set, for instance after a custom mass import or maybe when using direct S3 upload.
Accessing the content/length is going to access the blob to get its size, this means downloading the blob when the blob is stored in S3.
This is a known bottleneck during reindexing.
A workaround is to update the content/length to -1 at the database level
but it is better to have a specific procedure to populate this field with a correct value.
This could be a Bulk Action searching for a document with a blob and an empty or unset (-1) document/length.
Ideally, the value is set without downloading the blob and without triggering any listener (like a system property update).
This approach is better than having an option to avoid downloading the blob during indexing: