@federicomena you probably want to point people to bors-ng; it has a much more scalable algorithm (optimistic batches with bisection) https://bors.tech/