The pattern could have been play extended session, console MoBo and solder joints heat up and expand. System goes cold over three days (one day, any prolonged period of time), joints go cold, contract. Joints may contract to the point where all contacts are no longer creating a complete circuit.
Turn on system three days later and RROD appears.
In short, your console probably died as it was cooling down and all the components and connections were contracting.
It's the process of heating and contraction of the MoBo components, particularly the solder joints, as well as the design of the clamps of the heat sink which can physically warp the MoBo when shifting between long game sessions (high heat) and inactive sessions (back to cold) to a degree that when contracting while inactive, solder joints can break contact.
If the heatsink had been designed larger or more efficient from the beginning, heat could have been kept lower, preventing the MoBo warpage and near melting of solder joints. Misapplied thermal paste, as has been the case in some failures, can also lead to excessive heat, leading to the excessive expansion/contraction cycle and an inevitable RROD.
The other alternative, in addition to re-designing the X-clamp for the GPU heat sink, as well as the heat sink itself (larger or more efficient too keep the GPU cooler), it to simply make the GPU smaller (from 90nm to 65nm) and use less energy and create less heat. Falcon only made the CPU smaller, which has not been the primary source of failures. Yes, the 65nm CPU does produce less heat than a 90nm CPU, hence less heat inside the console under peak loads and an overall reduced failure rate (without directly addressing the real problem: the GPU). The Falcon is hardly a flawless solution.
Once all those solutions are implemented (65nm GPU/Jasper), it's just a matter of quality control to see that each redesigned console is manufactured to acceptable factory standards producing no more than a 3-5% failure rate.
