I agree, g.fast if done from poles would be a good advancement. But sadly I think once they realised they could still get the marketing benefits from a much cheaper cabinet only rollout, they went with that. I feel where they realise deeper fibre is needed they will just do FTTP.
Also I still believe profile 35 vectored bonded VDSL was the better choice than g.fast, look at what the germans are achieving with it. All they had to do was swap out line cards. Existing modems compatible with it and mature stable tech, also without rolling out deeper fibre.
Openreach seemed to pick the worst out of the 3 options.
1 - g.fast cabinet based, best cost to highest headline speed ratio, however improves only a fraction of vdsl capable lines. Speed can also drop vs vdsl.
2 - 35b bonded vectored VDSL, same low cost as cabinet based g.fast but will improve almost every circuit, often significantly if previous service was not vectored. Wont achieve as high headline speeds tho.
3 - node based g.fast, highest cost of the 3 solution as needs deeper fibre, but achieves both highest headline speeds and high % of vdsl circuit improvements of the 3 hybrid solutions.