I think Kitz hits the issue... BT realised they needed to overhaul their core network in order to support the traffic levels they saw coming. They obviously decided on a converged IP based core, rather than a telco core "bodged" into supporting ATM, though it probably involved some internal battles.
An IP based core changes the way you would backhaul MSANs, possibly even changes the MSANs you buy, so you need to choose carefully.
I suspect BT would have been faster if they didn't attempt to upgrade the voice network simultaneously.
As for the gradual rollout of WBC...
BT obviously completed the biggest exchanges earlier. The ones with space.
BTW forum presentations, relatively recently, were describing the process to upgrade exchanges with little space. This entailed adding 1 21CN MSAN, then forced upgrading off a 20CN DSLAM, then removal of the DSLAM hardware to create space for the next MSAN. And repeat...
Having reached something like 3600 or 3800 exchanges, they've got to the next phase ... the exchanges with no spare space at all.
Last week's ISP forum had BTW describe a new process that they're trialling for these exchanges. This entails the removal of a 20CN DSLAM, addition of a 20CN MSAN, and migration of all users in one overnight slot. While keeping telephony running in the meantime.
That's for 1,900 exchanges.
I suspect these kinds of forced migrations become a hefty overhead on both BT and ISPs.