Single-Shard Production E2E Tests: Improved/Innovative New Test Case

Step 1 — Baseline run

From the repository root (production tests are serial and use real ports):

RUST_LOG=info cargo test -p hyperscale-production --test runner runner_boots_listens_and_shuts_down_cleanly -- --test-threads=1 --nocapture

Confirm pass once. Note log lines for peer id, listen addresses, and clean shutdown.

Step 2 — Trace markers (what the harness can show)

Add three tracing::info! markers (preferred over println!):

1. After build() succeeds — e.g. info!("[lab] runner built");
2. After listen_addresses() returns — log addrs.len() and the addresses
3. Immediately before drop(shutdown) — e.g. info!("[lab] requesting shutdown");

Re-run with the same command. Verify ordering: built → listening → shutdown → test pass.

Step 3 — Assert observables the canonical test omits

After fetching listen addresses, add:

assert!(!addrs.is_empty(), "production runner should bind at least one QUIC listen address (validator libp2p mesh, not wallet HTTP)");
info!("[lab] listen addresses: {:?}", addrs);

This documents what production E2E can prove today: wiring + network surface, not yet tx execution.

Step 4 — User input beginning (design note, not required code)

In your journal (or PR description), answer in bullet form:

• Live path: HTTP client → submit_transaction_handler → tx_submission_tx.send(NodeInput::SubmitTransaction { ... }) → pinned loop on consensus_rx.
• Test hook: ProductionRunner::tx_submission_sender() sends the same NodeInput without RPC—useful for integration tests that need a tx without standing up HTTP.
• Why this test skips both: CI scope is build + libp2p + pinned loop + shutdown; adding a real signed tx is a follow-on test file, not a one-line tweak.

Stretch (optional): sketch a new #[serial] test that calls tx_submission_sender() after run() starts, with a fixture transaction from TestFixtures—do not block this lab on finishing that test.

Step 5 — Stretch: longer run window

Change the post-run() sleep from 500ms to 2s once, re-run, and note whether logs show periodic metrics or channel-depth exports (if enabled in your build). Restore 500ms before you finish so CI stays fast.