From a little watching of my connection activity, I observed that a lot of the upload bandwidth was loading older blocks. A limit on bandwidth provided for uploading older blocks would certainly help.
Implemented in Bitcoin Core git master a couple months ago, will be in 0.12; along with many other resource control tools. Though if widely used it will further congest getting blocks to begin with. It's a good trade-off, but not without costs. None of this is magic: cutting down this source of traffic or that is mostly just constant factor reductions. To operate the system needs a safety margin, and we're bumping up against actual limits.
It's good that instrumentation and controls are coming. IMO they are a couple years late.
If done right, controls would not slow down getting current blocks. These should get priority because the nodes requesting them are supporting the network. If nodes coming back on line or, worse, initially loading, get worse service at times of peak network load this is perfectly reasonable, IMO. (But I don't have any data, so I am forced to speculate.) Initial loading should be only from well connected nodes that source historical data. There is no reason why the real-time network should be forced to do a job of bulk data transmission for free. If this is a problem, there is no reason why these nodes couldn't charge for their service. For example, AWS charges about $0.10 USD per Gigabyte. It would be entirely reasonable to charge newbies a fee to offset these real costs. Or a business can sell preloaded bitcoin disks, like the 21 Bitcoin Computer.
As to constant factors. They are easily dismissed by theoreticians. Practical people who design, deploy, manage and operate businesses that provide cloud computing services live or die by small percentage changes in constant factors. In this case, it appears that there are tremendous opportunities for two types of improvements here. First, in reducing the amount of data that an operational node needs to transmit and receive, and second in tuning the nodes and their networking environment to effectively utilize the available resources.