The months of R&D to create a workaround could simply be because the subset of motherboards which trigger this issue are doing something borderline/unexpected with their voltage management, and finding a workaround for that behaviour in CPU microcode is non-trivial. Not all motherboard models appear to trigger the fault, which suggests that motherboard behaviour is at least a contributing factor to the problem.