Well, as an additional data point - this morning doing some basic work on my MBP but keeping the temperature sensor window open on my desktop I noticed three of the four CPU cores spike to 95 deg C, square in the yellow range, without the fans speeding up. Some people suggest resetting the SMC, which controls the fans, but it never seems to totally solve the problem. Seems that maybe the automatic fan control algorithm in the SMC has some flaws.
After a quick Google search I found another little tool that you can use to control the fans yourself, either by setting them to a constant RPM, or linking them to your own sensor algorithm. Will give this a try for the next few days and see if I can improve on it.
Anytime I had data on a panic, the CPU was in the red for temperature. And high power chips like the CPU and GPU have big heat sinks together with a heat conducting paste between the chip and heat sink. It's entirely possible that the whole issue about swapping boards is that once you go red, that paste may run to decrease its effectiveness making the board more vulnerable to future heavy use. But the root cause may very well be a bad heat management in the SMC that allows the system to get into a runaway state that it never can completely come back from. And even though it shows up as a GPU panic, it can be triggered by a CPU heat issue, and a GPU just reacting to this as the units coordinate execution paths, or nVidia chips being less heat tolerant than the Intel chips.
And as these kinds of design flaws go, they can propagate from generation to generation of system until someone throws that component out and starts all over, which may explain why we have multiple MBP generations which all seem to be plagued by the same problem.