The relevant information is the cost to the ISP. It's using that information that will help protect network neutrality. You can't scrape that from anywhere. What the torrent client does with that information is then the next issue. I can see that introducing an element of randomness may be useful.
Note that my suggesting is very simple to implement and costs nothing at run time (one failed or succeeded http request). If it turns out that the ISPs don't implement their side of it then that a) gives you an argument against them in the net neutrality debate and b) doesn't stop you going crazy with the heuristics.