Technical Question: obviously the speed of zk calculations is pretty important to the performance and throughput of Dusk... is it possible to use FPGAs to accelerate this, and if so, how much gain is there and what disadvantages are there?

It is very possible to accelerate zkp through FPGA. The drawback is a higher complexity which would limit decentralization. See, using FPGA on proving systems with rollups is desirable, since the proof is an infrastructural operation executed on a batch of transactions and is not a fundamental part of transactions which every user should generate. In our case instead, every user must generate a zero-knowledge proof when transacting, and therefore the inherent complexity would impact their capability to run a node independently