Status: OUT
The WAL's carefully engineered write-time durability infrastructure is systematically abandoned at recovery time: the two-tier durability model (per-write sync vs. batch-only fsync) loses all distinction during replay because recovery ignores tiers entirely, AND crash recovery is both broken (no safe path across any implementation) and unverified (no crash or async tests), meaning the engineering investment in write-time safety provides zero value when it is most needed.
Write-time durability engineering without matching recovery implementation is stranded investment; the most sophisticated half of the system is precisely the half that doesn't protect data.
Depends on (SL): wal-durability-tiers-lost-during-recovery, crash-recovery-both-broken-and-unverified