Thanks for posting a related issue! However, timeouts are not the case for me as nothing had timed out – everything was indeed extremely rapid and no timeout errors were reported by curl or any other component or logfile.
It is likely there is another issue causing the curl download errors in my case. One possibility may be extreme parallalism – if the download concurrency is set to, e.g. number of cores, it’s likely cdn is rejecting additional connections (I have 272 threads). If it is just a mirror refusing connection, then there may be a more graceful way of handling the problem. What are the chances a download could become corrupt (e.g. bad data is downloaded and resuming doesn’t fix the bad bits)? Is the final checksum tested against?
Thanks again.