Since most cabinets are Huawei and have had G.INP for a year they have plenty of information about what modems are and are not compatible with G.INP.
Things aren't as straightforward as that.
The standards allow for multiple implementations of G.INP, at both ends. When you see something working, all you can say is that one particular combination works.
With Huawei cabs working with lots of modems, we know that one set of combinations work - which might amount to around 10% of all combinations. But we know nothing about other combinations. So...
This huge mess is obviously down to a faulty and completely inadequately tested implementation of G.INP on the ECI cabinets.
... it is really hard to make this conclusion stick as the *only* possibility.
It is indeed one possibility amongst many. Unfortunately, faulty modem implementations are also a possibility - and become more likely if other modems work just fine.
What is certainly true is that the combination of ECI cabinet and at least one modem hasn't been tested together. Whose fault is that? Who specifies the range of modems that *should* be tested against? Should we expect more than, say, those modems put forward by ISPs for conformance testing?
IMO, there isn't one problem here, some aspects seem more likely to be caused by the cabinet; some by modems.