Sadly this also happens if both ends are running 9.1. The transfer stalled after about 1 GB of data having been transferred. After disabling SACK (on the local end) and restarting the transfer, it completed without any stalling (transferring another 9 GB of data).