There isn't an "error" / "bug" in either papers when one looks at them in isolation. Together, they seem to have a very subtle difference in the algorithm to trip most distributed systems practioners.