I used to be a advocate of anti buffer bloat, felt it was the biggest evil on internet performance metrics.
But when you hitting saturation conditions, something has to happen, the ISP could buffer the packets, so they still arrive, just delayed. This is buffer bloat. Or they could drop everything above your line rate (or the rate limit they set which might be below your line rate), this will lower latency under load, but I think the consequences are more severe, as you now dealing with higher levels of packet loss.
You are more likely to get buffer bloat (as well as packet loss) if the congestion window is excessive, sadly not a uncommon occurrence these days, if the download is multi threaded (speedtest.net is), if the RTT is lower (as this makes it easier to overshoot congestion window), I think BBR also makes it more likely on fat pipes.
I am pretty sure Zen use link aggregation, that would potentially cause jitter if the links are not consistent latency, but I wouldnt expect to have an impact on buffer bloat, the interesting thing about those results, is the one's with higher load latency are also slightly slower.
You think this is related to the Zen Manchester gateway?
The hop 1 might be in Manchester, hop 2 in London, Manchester in the name maybe because its indicating its the link to Manchester. I have seen hops named numerous times by ISPs by what they connecting to.
If everything is working still in all real world usage then thats the main thing that matters.
Finally, are all those speed tests from the same source? speedtest seems to love randomly selecting a server compared to using the same one every time.
