Transfers from VMware-virtualized Agents periodically fail with the error: The connection was aborted. An existing connection was forcibly closed by the remote host.
Although there are other possible causes of the transfer failure, this can be caused by using the (older) E1000 Virtual NIC on the Agent in question. Usually, this NIC type would be added to the VM when the VM is created, either manually or as part of a P2V conversion of a physical machine to a virtual machine. Using the E1000 Virtual NIC can cause bottlenecks in network performance on the ESX(i) Host/physical network that hotst the VM Agent. This NIC type is an older driver type (dating back to vmx-04 virtual hardware version, which was the standard in VMware ESX(i) 3.5. Newer virtual NIC types function much better (either the paravirtualized vmxnet3, or E1000E types,) and are preferred.
Change the VMs Virtual Hardware type (within VMware vCenter, or if no vCenter exists, on the VMWare host machine itself,) from E1000 to one of the preferred types (vmxnet3 or E1000E.) This will require a reboot of the guest OS, and a re-IP of the new NIC from within the OS on bootup.