Redis Sentinel nodes continuously monitor the master server. When a failure is detected by any Sentinel, it votes that the master is down. These votes are counted, and only if the number of votes reaches or exceeds the configured quorum does Sentinel start the failover process to elect a new master. If the votes are below quorum, Sentinels wait and continue monitoring. This quorum mechanism helps avoid false failovers caused by temporary network issues or isolated Sentinel failures. In the example, with quorum set to 2, failover starts only after two Sentinels agree the master is down.