I have a friend who is working in high-performance computing space. They have some workloads that they claim could be an ideal source for useful proof-of-work. Putting the discussion aside if the computations are suitable or not, she would like to test out plugging it in to a running blockchain.
Do any of Ethereum clients currently support pluggable proof-of-work algorithms e.g. assuming Bysantine Fault Tolerance consensus? What could be a starting point for such a research?
Not a complete answer, but Primecoin tried to provide useful p-o-w by letting miners search for primes of a certain size which is mathematically provable useful (here: number of digits).