Also, Bitcoin is kind of a "dirty" solution to the consensus problem; if all the academics were looking for an elegant solution it's not surprising that they didn't discover it.
It would be nice if science had a term for its equivalent of "outsider art."
Everyone keeping a complete history of all actions is, like, consensus 101.
Research is based around not having to do that.
This has already been shown to be untrue, an anonymous pool has mined both bitcoin cash and the segwit bitcoin chains.
That used to be a mounting problem in the Bitcoin world which began to fade away with the introduction of FPGA-based miners, but it's still a problem with a couple of altcoins. There are pros and cons with both compute-hard and memory-hard PoW but many seem to lean towards compute-hard being the less bad choice.
Do you have a source for this? I'm not even sure what CPU based currencies you are talking about. GPU based proof of work seems like even more of a stretch seeing as particular high end cards that are better with integers and bit shifting are usually used.
China is already well on their way to doing this, with Jihan Wu and the emergence of larger mining 'cartels'.
I also thought that Bitcoin's asic-vulnerability (and thus mining centralization) would be fatal. It turned out to be not fatal (yet), but thats not something which could be determined on paper. It needed real-world use before people knew if it could work or not.