For reference: Distributed Proof Generation
Some points that are not clear to me:
In general, when you are building a proof, you can reuse a huge percentage of the work to compute the proof, to generate a different proof with a different prover address. So, an obvious attack is to generate many proofs faking many provers, but reusing the work of generating one single proof.
How the price is calculated and adjusted? This price needs to be calculated/commited before hand, but may variate.
Generally speaking, for me, the bigest concern of a proving market is the danger that there is a single prover that ends up building all the proofs and geting all the fees, because it hase the more competitive advantage, and give no margin to the second one (the backup), that gets no fees, but need to maintain an infrastructure. This model can bring important outages in caise of main prover failure, because backaps are not economically worthy, so they just not exist.