2.4 Complete Protocol
The basic protocol maintains consistency, but it cannot ensure any progress because it states only what a priest may do; it does not require him to do anything. The complete protocol consists of the same six steps for conducting a ballot as the basic protocol. To help achieve progress, it includes the obvious additional requirement that priests perform steps 2–6 of the protocol as soon as possible. However, to meet the progress condition, it is necessary that some priest be required to perform step 1, which initiates a ballot. The key to the complete protocol lay in determining when a priest should initiate a ballot.LAMPORT, P. 12 — §2.4
The complete protocol isn't a recipe or an implementation but an exploration of the importance of leadership for liveness. Having a centralized leader to initiate the ballots simplifies the process and cuts down on chatter.
Never initiating a ballot will certainly prevent progress. However, initiating too many ballots can also prevent progress.LAMPORT, P. 12 — §2.4
The remainder of the section approaches the problem indirectly, using the parliamentary metaphor and a long discussion of time measured in minutes. If this feels unsatisfying, that reaction is reasonable. The paper fixates on delays and timing rules without providing a usable method for choosing them.
Still assuming that $p$ was the only priest initiating ballots, suppose that he were required to initiate a new ballot iff (i) he had not executed step 3 or step 5 within the previous 22 minutes, or (ii) he learned that another priest had initiated a higher-numbered ballot. If the Chamber doors were locked with $p$ and a majority set of priests inside, then a decree would be passed and recorded in the ledgers of all priests in the Chamber within 99 minutes. (It could take 22 minutes for p to start the next ballot, 22 more minutes to learn that another priest had initiated a larger-numbered ballot, then 55 minutes to complete steps 1–6 for a successful ballot.) Thus, the progress condition would be met if only a single priest, who did not leave the chamber, were initiating ballots.LAMPORT, P. 13 — §2.4
The specific numbers here do not matter, and the paper is not claiming that a system can or should measure time this way. The addition to the spec is simply this
If no one entered or left the Chamber, then after T minutes exactly one priest in the Chamber would consider himself to be the president.
The purpose of this section is not to define timeouts. It is to show that without leadership, the protocol can live lock, even though it remains safe. And how through leadership churn is diminished and progress is more assured.