If you mess up the database you can just reread the log files. It also helps manage backpressure during activity spikes where your db can't keep up.