Section 3.3.5 Dishonest Legislators and Honest Mistakes

Inconsistencies
Despite official assertions to the contrary, there must have been a few dishonest legislators in the history of Paxos. When caught, they were probably exiled. By sending contradictory messages, a malicious legislator could cause different legislators’ ledgers to be inconsistent. Inconsistency could also result from a lapse of memory by an honest legislator or messenger.
When inconsistencies were recognized, they could easily be corrected by passing decrees. For example, disagreement about the current olive tax could be eliminated by passing a new decree declaring the tax to have a certain value. The difficult problem lay in correcting inconsistent ledgers even if no one was aware of the inconsistency.
LAMPORT, P. 21 — §3.3.5

Inconsistencies themselves are not the hard problem, nor are their causes. The hard problem is correcting inconsistencies that persist without being detected. Detected inconsistencies can be corrected by passing a new decree that reasserts the correct value. The issue Lamport is concerned with is uncaught inconsistencies that reproduce and continue

Redundancy as a Solution
The existence of dishonesty or mistakes by legislators can be inferred from the redundant decrees that began appearing in ledgers several years after the founding of Parliament. For example, the decree 2605: The olive tax is 9 drachmas per ton was passed even though decree 2155 had already set the olive tax to 9 drachmas per ton, and no intervening decree had changed it. Parliament apparently cycled through its laws every six months so that even if legislators’ ledgers were initially inconsistent, all legislators would agree on the current law of the land within six months. It is believed that by the use of these redundant decrees, the Paxons made their Parliament self-stabilizing.
LAMPORT, P. 21 — §3.3.5

Even with Paxos safety guarantees, the system is not immune to lost state, forgotten decrees, or inconsistent ledgers.

Lamport admits this is unsatisfactory and that more research needs to be put to this idea.

For Further Study
Unfortunately, we don’t know exactly what sort of self-stabilization property the Paxon Parliament possessed or how it was achieved. Paxon mathematicians undoubtedly addressed the problem, but their work has not yet been found. I hope that future archaeological expeditions to Paxos will give high priority to the search for manuscripts on self-stabilization.
LAMPORT, P. 21 — §3.3.5