Congestion on only one protocol though? Maybe outside of BT's network but very unlikely to be VP/SVLAN due to that.
If anything, NNTP should be higher/more consistent if multithread.
The OP isn't reporting a problem on one protocol - and actually fails to tell us what NNTP behaviour is seen.
His report is that "latency increases to 1s" (presumably, but not necessarily, measured by a ping), and that "web browsing etc becomes impossible". NNTP causes this, but we don't know whether it is also a casualty.
To my mind, traffic shaping would be an option if NNTP were running slow, but not other protocols. A shaped NNTP would leave more room on the line for pings and browsing, not less.
In fact, it seems almost the opposite: that congestion exists, and NNTP should be managed, but isn't!
When the signs of congestion appear, what speed does NNTP run at? What speed are downloads then?
Are you comparing like with like? Is the same machine being used to generate the downloads as creates the NNTP request? Is it the same machine that experiences latency/browsing issues?
Does everything use the same LAN segment? Do you see latency/browsing issues via Ethernet, if the news transfer runs via WiFi? Or vice versa.