I just found a new PDF (or new to me, at least) with some predictions on behaviour of the 17a profile under different line conditions (ie different amounts of crosstalk).
It is a PDF from
Ikanos study into bonded VDSL2 solutions. The later detail discusses various bonded solutions, but it does start with some useful benchmark figures for unbonded circuits.
The graph on page 8 is the most enlightening: It shows how throughput degrades from a theoretical-best standalone circuit down to multiple co-interfering circuits, and suggests that BT's "up to 80Mbps" solution will be limited to about 300 metres from the cabinet.
Page 14 shows the downstream behaviour of the 12a profile - which ought to be very similar to the 8x profiles (the extra frequencies from 8MHz to 12MHz are only used for upstream).
UNFORTUNATELY...
The graphs are all based on the 998 band-plan, which is asymmetric, tilted towards giving more frequencies for downstream.
Openreach have been using band-plan 997 so far (with profile 8c), which is a more symmetric split between upstream & downstream. If they continue to use plan 997 with profile 17a, I'd guess that the downstream throughput will be lower than shown in the graphs.