@Jannes I don't really agree that there's any perceivable effect. There's upwards of 5000 peers you can download from in parallel and the client churns in an attempt to find fast peers. The bottleneck is literally never the speed at which you can download blocks, but the speed at which you can verify them. Even with blocks already on disk and a ridiculous amount of memory it is still a many hour, possibly multi day process to sync a new node on commodity hardware.
