I refuse to believe it takes a year to write a single firmware update, of course I am aware its possible there has been multiple updates written since the last public update last year, but we only know about the latest one.
So based on that I see 2 scenarios.
1 - ECI written an update to address the issues raised last year, and openreach sat on the update as they have a limited testing schedule, they started testing it this year prior to the expected march rollout and found a new problem they decided they cannot ignore.
2 - Several updates have been done but each keeps getting issues, and the latest one is the one andyhcz leaked. This however makes less sense as if they all kept having less problems why previously announce it was going ahead this spring?
I also believe when the rollout was underway with the M41s, BT were not even thinking of g.inp and vectoring, their concern was simply rolling out VDSL2 as quickly and cheaply as possible and its this reason I believe ECI have not broken their contract, BT are now simply asking them for some spec enhancement that wasnt originally thought about due to the short termism. Speculation on my part, based on what we know already. As someone else said this is probably a very low priority issue for BT, as g.inp does not affect the product they are marketing, it meets ASA requirements for a 80/20 product. The only business case for g.inp is that it can reduce fault reports as the technology is quite good at masking issues on lines.
The way to make you realise the situation is not as bad as it could have been is that without the cheaper ECI cabinets I expect the commercial rollout would have covered less properties, and I certainly prefer a ECI based VDSL service over either ADSL or cable broadband.
What about the viability of running a 3db target without g.inp? its viable. As shown on ADSL but I am not sure if they will risk it as I think fault reports will inevitably increase. My own opinion is that the best boundaries for SNRM targets is 1,4,7,10 db and so on not 3,6,9. My own line has comparable error rates with a 4.5db snrm vs a 6db snrm. If my theory is right it would increase when dropping below 4db but anything above 4db it will have a comparable ES rate to 6db.