Good post, Kitz.
I missed it the first few times around ... I seem to have a blindness for things in the sticky section - maybe the blue "New" icon just doesn't stand out against the bold blue text of the title and the darker blue background!
Its an interesting notion that the ECI modems will have gone through one firmware upgrade to go from <totally incompatible> to <quite incompatible>, but it certainly seems to be true that a lot are waiting for something more. I had originally thought that it was a problem with the agent failing to update some modems, but this idea probably fits the evidence better.
Have we seen anyone with an ECI modem that appears to have had G.INP activated with no ill effects? Any hint, anywhere? With ECI's being considerably harder to unlock, I imagine we will have a hard time confirming it ... other than the kind of efforts going on via TP-Link.
I believe someone suggested that US rate of 20000 was synonoomommous with g.inp?
Ive always sync'd at 20,000 upstream. Ive never paid much attention to the 79999, 79998 or 79987 etc as I thought it was perhaps just a quirk of the modem/dslam.
These are some of mine
79999 20000 HG612
79987 20000 Zyxel
80000 20000 TP-Link
I thought exactly the same. I bet it was w3 that pointed it out too 
Not me.
I've always thought that the sync ending up a few Kbps short was a quirk - perhaps of the specs, perhaps of the chipsets. My best guess surrounds the algorithm that decides on bit-loading - that it just stops when the next addition would take you over the limit, leaving you on or just under. It might be even worse if following some water-filling technique.
Having seen this quirk happen often enough, I wouldn't bother looking for non-quirks in G.INP. Until recently, there have been too few reports to draw real conclusions.