OK so this is an unusual one, and one I've been having a really hard time explaining to VM's tech support - about the only one to show some understanding was a guy in second-level tech support who agreed there was a problem somewhere after I demonstrated via a LogMeIn session, but basically said the onus is on me to prove it's VM at fault before they'll be able to elevate it to the networks team.
Basically, single TCP stream speed is terrible - we're on 200Mb and a single e.g. HTTP/FTP download won't go past about 8-10Mb most of the time, getting worse at peak times (I'm not confusing b and B BTW). However, with Speedtest using 20+ parallel streams to test, it of course doesn't show up this issue. However Thinkbroadband's speed tester which does separate x1 and x6 stream tests, does show it (link), but conveniently VM won't accept it as a recognised test.
Further to this, I've asked a few friends to do similar tests and those on VM but in other cities are unaffacted, and people on BT/Sky/etc are also unaffacted, even those nearby. However a nearby friend on VM has the same issue just to show it's not specific to me (I don't think it's a last-mile congestion issue either).
Anyway, I wonder if anyone has any ideas, or if a few people would mind doing similar tests (e.g. try downloading Linux isos or that TBB speed test) to see if any patterns emerge? The first two hops of my traceroutes are the Knowsley CMTS, then Preston core. Annoyingly the next couple of hops now drop ICMP but they used to be Manchester and Leeds for most routes IIRC.
Thanks!!