I've speculated that maybe it's a bit of both. To sum up: the lighting is controlled by a dedicated microcontroller that talks on a serial bus called I2C. The microcontroller has RAM and a ROM burned with the program to do the lighting effects.
What I think may have happened is that there is a flaw in the microcontroller that has resulted in some kind of data corruption that freezes or crashes the program. If there isn't a way to reinitialize that controller's data, it may not be recoverable without an RMA. Not the same thing quite as a hardware failure, just some code on the lighting controller that has an unanticipated path to failure.
What has confused the issue is another RGB problem involving RGB RAM (which I don't have) which creates a problem when the lighting control software and other "direct hardware access" type of programs conflict. What is needed is what's called a "mutex" or "mutual exclusion" object that acts as a go between and prevents conflicts. This is a common approach when programs must both access the same hardware. It was not done for Aura and the other software that may be conflicting with Aura and may explain the memory problems as well. The problem can be so severe that the SPD data on DIMM's becomes corrupted and the memory ceases to function until the SPD is rewritten with good data.
So, a host of problems involving the lighting and the fact that we may have one common symptom for multiple causes is hamstringing getting it definitively fixed. Some folks have found short term workarounds that seem to fix it but it eventually breaks again. I've had some arguments on this very board with some folks who claim to have "solved the problem" but the solution usually involved the digital equivalent of clicking your ruby slippers together and chanting "there's not board like ROG, there's no board like ROG" until you get back to Kansas...
Tired of trolls and mods that act like this platform has no problems and it's the users fault. Later.