How to generate SNARK parameters securely

The messages that have to be passed from one node in the Secure Multiparty Computation to the next node are on the order of gigabytes in size. And, if any node fails (stops responding), the rest of the nodes can’t make progress without it, so we’d have to kick that unresponsive node out of the Secure Multiparty Computation set and restart the protocol.

So, for a bunch of practical operational reasons we can’t handle lots of participants.