I'll leave this as an open question. 

This design is a bit different from azure-storage-blob where chunked download is only leveraged when performing download to file operations as disk writing should support non-sequential writes. Or, better put if we concurrently download ranges 0-4MB and 12MB-16MB, and before 4-8MB and 8MB-12MB complete, this is fine as the file ranges can still be written correctly.

This however is a case of we're downloading an unknown size container, which in this case azure-storage-blob uses a concept of open-ended download, Range: start- which indicates download from a given start point until completion. The download is then tracked for progression, such as 12MB has been downloaded, and resuming logic is added on top if there is a failure, so if at 12MB a network error happens another download call is made with Range: progress- to resume the download until either completion or exhaustion of retries (such as a complete networking failure).

To me this should be following the latter logic.

Originally posted by @alzimmermsft in #33348 (comment)