Status: OUT
The WAL's two-tier durability model provides effective crash protection: critical operations (checkpoints, batch commits, rotations) unconditionally force fsync while per-write sync respects the configured mode, creating a meaningful durability hierarchy where batch boundaries are always durable regardless of the performance/durability tradeoff chosen for individual writes.
Two depth-1 conclusions about WAL durability form a coherent positive model, gated on the macOS fsync limitation — on Darwin, os.fsync() may not flush the disk write cache, undermining the entire durability hierarchy on the development platform
Depends on (SL): wal-two-tier-durability-model, critical-wal-operations-always-force-fsync