Worst case for the probabalistic finality due to subsampling

Hey guys, a huge “abacs” fan here hehe. I have a question with regards to the consensus and more precisely what would happen if that extremely small probability that consensus is not reached happens? We know that the Avalanche consensus sacrifices this property to achieve a scalable consensus that can accomodate thousands of validators without hurting the performance (because of subsampling), but let’s assume a worst case scenario. What would be the consequences in that case for the network? What would be the best course of action then and would that mean that only liveness would suffer or? Just curious about such edge cases that are highly improbable but still have a probability > 0.

1 Like

If there’s some kind of liveness attack or network issue then, well, the network loses liveness. No blocks will be accepted. Eventually it will recover and it’ll proceed from there. The consensus engine will continuously try to keep making progress. It’s not like there comes a point where it says “well, we’re stuck” and just stops. Everything will eventually be decided.


Awesome, just wanted to be sure whether liveness takes the damage, thanks!