fix(storage): propagate clock error in applied_at_ms (TD-08)

[intendednull]

what

migration insert in crates/storage/src/store.rs did:

SystemTime::now().duration_since(UNIX_EPOCH)
    .map(|d| d.as_millis() as i64)
    .unwrap_or(0);

clock pre-1970 -> applied_at_ms = 0 silently lands in schema_version. rest of crate hands back anyhow::Result. mismatch.

fix

.map_err(anyhow::Error::from)? on the duration_since. function already returns anyhow::Result<()>, the ? just falls through.

only one call site in the storage crate (grep clean). no shared helper, one line plus comment.

test

extended schema_version_table_exists_after_open. now reads back applied_at_ms alongside version and asserts every row has applied_at_ms > 0. pins behaviour off the legacy 0 fallback under normal clock.

error-branch test (clock pre-1970) would need SystemTime mocking infra the crate does not have today. out of scope for this issue. comment in test notes that.

storage tier (lowest covering this insert path).

verify

$ cargo fmt --check -p willow-storage
(clean)

$ cargo clippy -p willow-storage --all-targets -- -D warnings
    Finished `dev` profile [unoptimized + debuginfo] target(s) in 0.40s

$ cargo test -p willow-storage
test result: ok. 35 passed; 0 failed; 0 ignored; 0 measured

storage is native-only (rusqlite + tokio in Cargo.toml), no wasm check.

Refs #254

---

Generated by Claude Code