MonadBFT: Fast, Responsive, Fork-Resistant Streamlined Consensus
Mohammad Jalalzai
Session summary
The presenter introduces MonadBFT as a consensus protocol built on pipelined, two‑chain BFT designs in the HotStuff family, aiming to keep leader rotation, pipelining, linear message complexity, and optimistic responsiveness. While reviewing standard protocols, the presenter notes practical concerns observed when considering deployment: speculative finality based only on seeing a quorum certificate can be unsafe under partial delivery, valid blocks can be silently abandoned (“tail forking”), and a single faulty leader may cause multiple consecutive timeouts.
To address these issues, the authors add mechanisms while preserving the underlying model of Byzantine faults under partial synchrony. MonadBFT introduces a “tip” (a proposal without payload) that validators attach to timeout messages so later leaders re-propose blocks that already gathered votes, mitigating tail forking and strengthening speculative finality except in cases of leader equivocation. Validators also send votes to both the current and next leader, allowing the current leader to propagate quorum certificates and confining failures to a single round. The presenter reports that this protocol has been running on the Monad mainnet for several months.
Key takeaways
- The authors identify speculative finality violations, tail forking, and multi-timeout failures as practical issues in existing pipelined BFT protocols they evaluated for deployment.
- MonadBFT introduces a “tip” included in timeout messages so future leaders must re-propose the highest-view tipped block when it is fresher than any quorum certificate, reducing silent abandonment of honest proposals.
- Validators in MonadBFT send votes to both current and next leaders, so a single leader failure causes at most one timeout, improving responsiveness while retaining HotStuff-style properties.