Research Note · Consensus

How BFT Consensus Actually Reaches Finality

By Dr. Elena Muster, Protocol Research · September 18, 2026 · 8 min read

Every introduction to Byzantine fault tolerant (BFT) consensus promises the same thing: once a block is "final," it will never be reversed, even if up to a third of participants are malicious. The promise sounds like magic until you see the one geometric fact underneath it. This note explains that fact without heavy math.

The setup

Imagine n validators, of which at most f are dishonest. Classic BFT protocols require n ≥ 3f + 1, and a decision requires a quorum of 2f + 1 votes. Two numbers do all the work:

The key insight: quorums always intersect

Take any two quorums of size 2f + 1 drawn from 3f + 1 validators. Suppose they did not overlap. Then together they would contain at least 4f + 2 validators — but only 3f + 1 exist. Contradiction. Therefore any two quorums share at least one validator, and that shared validator cannot be in two places at once: it cannot honestly vote for two different blocks in the same round.

"Finality is not a property of signatures or speed. It is a property of set arithmetic: two big enough groups in a small enough room must overlap."

If the shared validator is honest, it votes once, so at most one value can gather a quorum. If it is dishonest, it may double-vote — but then it gets caught, because two conflicting quorum certificates expose its misbehavior with cryptographic proof. Either way, the protocol wins: conflicting finality is impossible, and misbehavior is attributable.

Why f = n/3 is the ceiling

Could we tolerate more than a third dishonest? Only if quorums could intersect in honest members, which requires quorums larger than two-thirds — but then two quorums no longer fit in the validator set (2 × ⅔ = &frac43; > 1). This is not an engineering limit to be optimized away; it is the same 1982 result from Lamport, Shostak and Pease, and no message-passing protocol can do better in the asynchronous model.

What "finality" does and does not mean

Further reading

The original Byzantine Generals result (Lamport, Shostak, Pease, 1982), Castro and Liskov's PBFT (1999), and our own preprint on committee randomness all build directly on the quorum-intersection argument sketched here.

This note is educational content about distributed systems. It is not investment, legal, or financial advice, and it discusses no financial products. See our Independence Notice.

← All research notes