Dear Leander,
Thank you very much for your interest in that paper. All our implementation with regard to collective signing is here:
We did never get around to handle a witness failing _after_ the commit, only before. What we currently do, if that happens, is to restart the protocol. As we use the CoSi protocol with our industrial partners, the most common fail-case is a node that is down, or too busy. So restarting is fine.
We have, however, some work in progress here:
Which is currently completely unsupported and a mere student project. The idea is to replace the schnorr-signatures with BLS signatures, then we don’t need two rounds, and don’t have the problem of failing nodes after the 1st round.
What is your use-case?
Sincerely,