After an async createReplica job on a feature service finishes, the replica is ready to download using a link given on the job status page. One can download the replica all at once. But, like most REST endpoints, one could use the common HTTP range header to download the replica a bit faster:
https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Range
For example, the replica could be downloaded in 10MB chunks, 5 chunks at a time. Then, the chunks would be reassembled programmatically after all chunks were downloaded. This is a fairly common practice for larger HTTP downloads. From my tests, ArcGIS Server does seem to respond properly when called with the RANGE header (replies with 206, gives back only the number of bytes requested). But, I'd like official support if we are to base code on this behavior. Is it officially supported?
A few examples of others using the range header:
https://stackoverflow.com/questions/3303029/http-range-header
https://www.infoworld.com/article/3198668/how-to-work-with-http-range-headers-in-webapi.html