I think the PPP issue has been horribly misdiagnosed.
g.inp shouldnt break the protocol, the modem is a bridge passing unmodified packets across, I still think these PPP issues are because of high amounts of loss of sync causing ghosted PPP sessions and end users having to wait for those sessions to timeout.
Whatever the cause, it patently isn't simple, else we'd have seen it happen with the Huawei activation.
Some of the symptoms sound, to me, a little like he software in the CPU is losing control of the hardware, suggesting exceptions and timeouts are happening. Lots of fun to investigate.
Other symptoms are, frankly bizarre. High FEC rates for a precise 1 hour period? Doesn't sound real, and seems more likely to be a reporting issue. Memory references going wrong?
Interworking with other vendor's equipment has, at times, been the bane of my life.