Interesting, thanks for the follow up!
I guess that makes a certain amount of sense, long unshielded cables often pick up RF and I guess in a data cable it might read as a stream of bad requests. I'd be interested to know the theoretical impact, in terms of how many legitimate requests are expected to be below-threshold. If they're only installed on 'noisey' lines, maybe the answer is "none in practice", since any packet that would be filtered would also be wiped out by RF interference?
I'm stunned to hear that you've consistently gotten over advertised rates. I've had Comcast in two places, and considered myself lucky when I got close to advertised. One apartment ranged from 80%-105% of advertised, then plummeted to 20% in the last few months I was there. (No troubleshooting available, because I didn't rent my modem from them.) The other just sat at 20% non-stop, except that supposedly Comcast's speed test server was getting my full allocated speed. Honestly, I was halfway convinced by the theory that ISPs throttle connections just so they can report annual "improvements" to consumers and the FTC. But they legally defend those advertised speeds as averages, so I guess someone has to be over the rating to offset the people who aren't.
...now that I think about it, could those "average download speeds" be from a mix of fiber-to-home and fiber-to-hub residences? If they're the mean of something bimodal, it'd explain a lot of what I hear.