fix(cdp): keep default raw readiness best effort

[aram-devdocs]

Summary

• keep explicit raw --wait-for selectors strict
• make the synthetic raw body readiness probe best-effort before DOM snapshot capture
• preserves the raw fallback path while avoiding the macOS CI failure where wait_for_selector "body" exhausted its 10s budget after a page-init stall

Validation

• cargo fmt --all -- --check
• cargo test -p plumb-cdp --lib raw_
• cargo clippy -p plumb-cdp -p plumb-cli --all-targets --all-features -- -D warnings
• cargo build --release -p plumb-cli
• cargo run --release -p plumb-e2e -- --site html-css --site tailwind-html --site react-vite --no-build --plumb-bin target/release/plumb --chrome-path "/Applications/Google Chrome.app/Contents/MacOS/Google Chrome"

[github-actions]

I have everything I need. Here is the full review.

---

PR #310 — fix(cdp): keep default raw readiness best effort

Scope: crates/plumb-cdp/src/lib.rs only (+24 / -10). Single-crate, no public API additions, no dependency changes.

---

1. Determinism

No violations. The change touches only CDP I/O orchestration; nothing flows into PlumbSnapshot output. tokio::time::sleep already exists in wait_for_selector_raw — no new wall-clock source. tracing::debug! is the correct I/O (not println!/eprintln!), consistent with plumb-cdp CLAUDE.md.

2. Workspace layering

Clean. All changes stay inside plumb-cdp. No new cross-crate dependencies, no new unsafe blocks, no println!/eprintln! added.

3. Error handling

The semantic split is sound:

• Explicit user selector (wait_for_selector set) → strict: error propagated with ?. Breaking a user's explicit wait contract is always fatal.
• Default body probe → best-effort: error swallowed after tracing::debug!. A macOS page-init stall should not abort the whole capture for a probe that was already a heuristic.

wait_for_default_ready_selector_raw_best_effort returns () — the name makes the contract explicit and the log line preserves observability for debugging. No unwrap/expect anywhere.

4. Test coverage

The two renamed unit tests (raw_strict_ready_selector_is_absent_without_user_selector, raw_strict_ready_selector_preserves_user_selector) correctly cover the refactored raw_strict_ready_selector helper. wait_for_default_ready_selector_raw_best_effort has no direct unit test, but it's an async function that requires a live RawCdpClient/RawPage — mocking would violate Plumb's "don't mock CDP I/O" convention. The e2e run cited in the PR description (--site html-css --site tailwind-html --site react-vite) is the appropriate coverage tier.

5. Documentation

All modified/added functions are private. missing_docs does not apply. No public items changed.

---

Punch list

No blockers or warnings found. The change is a minimal, well-scoped behavioural fix with the semantics clearly encoded in the function names and the log message.

Verdict: APPROVE