Checkpoints reduce recovery time by flushing dirty pages to disk periodically.
Without checkpoints:
Recovery replays entire WAL from the beginning. Could take hours for large databases.
With checkpoints:
Periodically write all dirty pages to disk
Record checkpoint position in WAL
Recovery only replays from last checkpoint
Trade-off:
Frequent checkpoints speed up recovery but cause I/O spikes. Configure based on acceptable recovery time vs performance impact.