Right now the next step would be enabling multithreading with Puma. Throughput is acceptable, packaging with Omnibus is working well. So the problem to solve is memory usage. We hope multithreading can help there.

Their blocker was a C extension I think, but with TruffleC that worry should be gone.

I'm personally looking forward Truffle implementation just because of that (and also benchmarks are looking really nice).

I don't think this will be out anytime soon TBH.