General questions about write ahead logs and recovery

I've recently learned about write ahead logs and failure recovery in the context of key-value stores like Cassandra and have some follow-up questions: 1. Do WALs get persisted into smaller snapshots so that we only need to recover from the latest snapshot instead of the entire commit history? What is the high-level mechanism for persisting and recovering from these snapshots? 2. Do logs get batched when writing to disk for performance? Is this at the risk of permanently losing data? 3. What happens when a machine fails during recovery from a WAL? Do we end up with partial results in the database? How do we recover from this state the next time the DB starts?