Handoff: Count final Actions first attempts for Ping cleanup

RE-VERIFY: Written 2026-10-28 17:41 UTC. Final GraphQL read confirms the PR OPEN at 5a03a90cfab4ec344c39c1c92643ac969ff3bdd3 and required Actions bld-build IN_PROGRESS: https://github.com/CodexCoder21Organization/UrlResolver/actions/runs/40731398401 . Full Actions XML and first-attempt counts remain unavailable. The bounded plain watcher stopped by its exact PID at 17:35 UTC, exited 130, and its PID and wrapper are gone. Recheck https://github.com/CodexCoder21Organization/UrlResolver/pull/1610 through GraphQL and its exact-head Actions XML. A fresh result wins over this document. GraphQL verified source/checkpoint/evidence refs at 17:41 UTC.

Design decisions in force

The 2026-10-28 FIX1610b brief says: "Never merge, enqueue, deploy or publish. Rebase only on a real conflict. Poll GitHub at most once every 4 minutes, using GraphQL. Keep notes short and plain." It also says: "Do not chase (3) here. A separate lane is investigating repeated one-off timeouts across UrlResolver CI." Keep every assertion; never raise timeouts/limits, reduce iterations, tune the JVM, use reflection/mocks, skip tests, or rerun CI unchanged. Use targeted local tests and plain build-watchman --pr, with a tracked process and deadline; stop every owned watcher before ending.

Mission and verified work

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 and "...super out of date when it should be fast and reliable". The original LIBP2P lane opened the Ping cleanup PR; REV1610 found an address-state leak, unnecessary broader setup, and an unrelated CI timeout. FIX1610b was asked to fix the leak test-first, resolve the setup scope, then report final Actions first-attempt counts within 60 minutes.

1. Both Ping finally blocks called setExplicitPublicIp(previous) only for non-null values. clear() clears observed addresses, leaving the explicit setting at 127.0.0.1. Temporary standalone copies of the actual tests force an assertion after PING_DECISION, while publication/settlement remain held, and check the exact public getter after finally returns. Both fail 3/3 baseline attempts: expected null, actual 127.0.0.1. Native XML contains all six failures and full traces.
2. The fix clears observations, restores the nullable prior value, and asserts the exact restored IP as three independent cleanup actions. Cleanup still retains the original failure and suppressed errors. Each forced-failure wrapper checks saved null and 192.0.2.17 in the same JVM, exact failure-object identity, and zero suppressed cleanup failures.
3. Four single-attempt controls showed both original tests pass without either the added duplicate negotiation gate or the explicit duplicate caller close on pinned snapshot-32. Those additions were removed rather than justified. The final combined setup passed both original tests and both forced-failure wrappers: 4/4 first attempts with maxTestAttempts=1 and no retry histories. Every original assertion remains in order: main 19/19 and 21/21; reviewed head 27/27 and 29/29. No budgets changed.
4. GraphQL confirmed OPEN immediately before each PR push. Final source is 5a03a90cfab4ec344c39c1c92643ac969ff3bdd3; the PR description gives the provenance, mechanism, removal controls, local counts and evidence links. Only the two original tests differ. Temporary wrappers/controls are evidence only. Source, documentation and plan checkouts are clean, with no stashes or local-only commits.

Relevant PRs and refs

Review evidence: https://github.com/CodexCoder21Organization/UrlResolver/blob/wip/REV1610-evidence/findings.md . Final lane findings: https://github.com/CodexCoder21Organization/UrlResolver/blob/wip/FIX1610b-evidence/findings.md . Exact reproduction commands: https://github.com/CodexCoder21Organization/UrlResolver/blob/wip/FIX1610b-evidence/FIX1610b/reproduce.md . Final XML: https://github.com/CodexCoder21Organization/UrlResolver/blob/wip/FIX1610b-evidence/FIX1610b/final-single.xml .

Deployed or published but unmerged: none. The lane's action audit includes only source/evidence pushes, PR-description updates, read-only GitHub monitoring and handoff metadata. No production access, deployments, publications, restarts or route changes occurred. Build outputs, dependency caches and machine-specific classpaths are deliberately omitted; they regenerate from the checked-in build scripts. The earlier separate release recommendation remains linked from the LIBP2P evidence branch and is outside this lane.

Remaining steps

1. Re-verify current PR/head first using GraphQL, at least four minutes after any previous status poll. Keep the no-merge/no-enqueue/no-deploy/no-publish instruction.
2. Watch only the PR's Actions gate with coursier launch buildwatchman:build-watchman:0.0.21 -r https://kotlin.directory -- --repo CodexCoder21Organization/UrlResolver --pr 1610, in a tracked background task with a deadline. Required bld-build is still pending at this snapshot; the optional kotlin-build-ci-test suite has zero dispatched runs and was not re-requested. Do not start or wait on remote buildtest suites.
3. Once Actions finishes, download the exact final-head test-attempt-records artifact. Use FIX1610b/count-first-attempts.py on its native XML, honoring nested attempt-history. Count attempt-2/3 passes as first-attempt failures. Verify both changed tests specifically, report all first-attempt counts and retain the artifact/counts in the evidence branch. Update the PR description with those counts.
4. Leave the shared one-off timeout investigation with its separate lane. No CI reruns to seek a pass. Report the census to the orchestrator, stop the watcher and complete this handoff once its Actions audit is finished.

Operational notes

A GitHub object receiver rejected one topic push with remote unpack/index-pack errors; the identical push succeeded on the next try. The plain record is in evidence. The report-challenge tool was not invoked because it automatically enqueues/merges. The runner's CLI has no single-attempt option, so evidence/SingleAttempt.java calls the public WorkspaceKt.execute API with maxTestAttempts=1 and lets the runner write native XML; it adds no JVM tuning. Testing guidance was read at https://github.com/CodexCoder21Organization/DocumentationRepository/blob/main/architecture/TESTING.md and PHILOSOPHY.md, plus the repository README and KotlinBuild instructions.