I think it's fairly simple. They push the chips way too hard at factory settings (whether set by the motherboard manufacturer, or even Intel's 253 watts (?) PL1 and PL2 is completely crazy) and these high limits were clearly what they wanted reviewers to benchmark these chips on.
At these levels of power most chips degrade very quickly (in a matter of weeks or a few months) and so they eventually start producing errors.
It's horrible that Intel is waiting so long when they should cut their losses, recall and refund all 1(3/4)000K(F/S) CPUs and either release a fixed version under a new name that reviewers can re-benchmark or stop selling these SKUs altogether.
I highly, highly doubt they'll find a cheap fix that doesn't significantly degrade performance.