Handoff: Finish the first-attempt CI audit for UrlResolver relay verification

Written 2026-10-28 16:53 UTC. RE-VERIFY: this audit is complete, but branch and CI state can change. GraphQL confirmed https://github.com/CodexCoder21Organization/UrlResolver/pull/1584 OPEN/MERGEABLE/CLEAN at d5e2fd6183e2547e89696a393ec08340ac9f4764 at 16:51 UTC.

Directions in force

2026-10-28: “Never merge, enqueue, deploy or publish.” “Rebase only on a real conflict; if GitHub reports BEHIND with no conflict, do not rebase.” No raised timeouts or limits, reduced iterations, weaker or removed assertions, reflection, mocks or JVM tuning. GitHub polling is at most once every four minutes. Evidence remains outside the PR diff.

Mission and result

The owner reported “I feel like https://droplets.wasmserver.com/ is super unreliable ...” and “...super out of date when it should be fast and reliable”. CEN1584b completed the requested first-attempt CI audit and returns READY for the relay-route change. No merge was requested or performed.

Findings

1. Fresh Actions https://github.com/CodexCoder21Organization/UrlResolver/actions/runs/40711098683 succeeded on run attempt 1. Its single XML has 3,141 tests, 3,139 first-attempt passes, two retried tests, two extra attempts, 3,143 total attempts and no final failures. All nine added relay scenarios passed first attempt.
2. Actions checked out f3ec602606a6442d21ca2cb0396b9886bb0b692e, the GitHub merge of main 0377f7468a5a713e3cf6a0f1b3fd203206964da9 and PR d5e2fd6. Its inventory exactly equals that main's 3,132 tests plus nine PR tests. The prior 3,054-only inventory assumption was refuted; both the stopped diagnostic and corrected inventory are retained.
3. Fresh testActiveQueryExhaustsTargetsAfterOpenAndReadFailures has the exact reused-address failure on the main parent in https://github.com/CodexCoder21Organization/UrlResolver/actions/runs/40709768271 . Its active-query source is unchanged and relay is disabled.
4. Fresh testReviewInitialPersistentProviderIsNotNoProvider failed its second direct call at the existing 1,000 ms native connection deadline. A real accepted TCP handshake held before completion forces the same second-call exception chain on main and head, 1/1 each without retries. Native controls pass 1/1 each. Four exercised functions match main parent, merge and head exactly; relay fallback is disabled. The natural reason the healthy CI handshake missed its deadline remains unexplained. This audit demonstrates shared direct-timeout behavior and no PR-specific difference; it does not claim to fix that separate behavior.
5. Original Actions https://github.com/CodexCoder21Organization/UrlResolver/actions/runs/40523067676 had 3,051/3,054 first-attempt passes, three retried tests and four extra attempts. All three are now proven pre-existing. Main https://github.com/CodexCoder21Organization/UrlResolver/actions/runs/40518327381 has both the exact relay-shutdown timeout assertion and withdrawal failure. A deterministic diagnostic-writer/Clock gate forces the same interrupted-worker cleanup aggregate on head and main, 1/1 each; main already corrects that test via https://github.com/CodexCoder21Organization/UrlResolver/pull/1586 .
6. Read every failed test and exercised code; source hashes and complete stacks are preserved. Fourteen main XMLs are counted separately. The prior local 12/12 gate and eight relay-only proof passes remain on the earlier evidence branch. No PR-caused defect was confirmed, so no product fix was made.
7. The plain watcher used PID 26661 and a parent Popen.wait(timeout=3600), default 240-second cadence. It exited GREEN/0 at 16:42:11 UTC and is gone. Temporary local probes were removed after saving their source and XML. Source checkouts are clean; no stashes or unpushed source work remain. No remote suite was manually started or awaited.

Refs

No code was merged, enqueued, deployed or published by this lane. No production access was made. Build outputs and repetitive compiler warnings are excluded; XML retains complete failure traces. Evidence branches have no PR.

Remaining work

No audit work remains. The broader suite still has disclosed existing failures and the unexplained natural direct-handshake delay. No production fix for those separate paths is claimed. Merging or deploying this PR is outside this completed brief and remains forbidden without a new instruction.

Reusable details

Use audit_xml.py for every runner XML: final success flags hide nested retry histories. summarize_fresh.py validates the exact merge inventory from main-parent XML plus the nine new scenarios. Actions checkout can use a GitHub merge even when run metadata names the PR head. Save that resolved checkout and its parents before comparing inventories. SingleAttemptCli.java/install_single_attempt.py wrap the public workspace with maxAttempts=1 in an ignored launcher while preserving scripts/test.bash --local --test <name> --log <outside-repo.xml>. worker-ordering-probe.kts and direct-handshake-probe.kts keep the deterministic comparisons; they retain existing assertions and budgets and belong only on the evidence branch. The direct-handshake probe forces a deadline condition, not the unknown natural reason a healthy CI connection missed it.