std::fs::File::sync_all durability guarantee for backup restoration verification

Portable durability verification after backup restore

Restoring backup data requires confidence that sync_all has flushed both file content and metadata to non‑volatile storage. The standard library documents sync_all as a best‑effort operation whose exact guarantee depends on the underlying OS and filesystem. On Linux it maps to fsync(2), on Windows to FlushFileBuffers, and on macOS to fcntl(F_FULLFSYNC) when available. However, drive write caches enabled by default on many consumer SSDs and HDDs can still lose data on power loss, and copy‑on‑write filesystems such as ZFS or btrfs introduce additional semantics.

No portable Rust API exists to confirm that restored bytes have reached persistent media. Applications must either accept the platform‑specific behavior or depend on external crates (fsync, durable) that expose stronger abstractions. The missing constraint is a documented minimum durability level that sync_all provides across supported targets.

Open questions

• Should std::fs expose a portable sync_durable (or similar) method that includes cache‑flush semantics, or is the current sync_all contract intentionally minimal?
• What is the recommended verification strategy for backup restoration workflows that must survive sudden power loss on Linux, Windows, and macOS?