Wrote up the WAL design and crash-test harness in detail
Published a deep dive on how KVStore's write-ahead log actually works — the CRC32C-checksummed record format, group commit, the two durability policies (always / everysec), and the crash-test harness that kills the process with kill -9 at a random point mid-write, 24 times across both policies, to prove recovery actually holds up.
Full writeup: raqueeb.hashnode.dev/building-a-crash-safe-write-ahead-log-in-go-and-the-24-ways-i-tried-to-break-it Case study: raqueeb.dev/projects/kvstore