Jul.IA v1.32.1-rc.1 — release candidate

Jul.IA v1.32.1-rc.1

Published release candidate. It validates the selected
correction tranche and the complete artifact path. It is not the stable
release and does not claim completion of work outside the tranche.

Immutable source and execution evidence

• Tag and source SHA: → 9a936d0
• Exact final-main CI: — success
• Release workflow: — preflight, full-tag gate, five-minute
  release soak-smoke, 12-cell build matrix, attestations, and draft publication
  all succeeded
• Release state: draft + prerelease + unpublished

Delivered corrections in this candidate

• strict rejection of unknown TOML fields and fail-closed validation of invalid
  known configuration values (#123);
• explicit default-on/disabled access-log semantics, dormant validated sinks,
  truthful restart classification, and ignored deprecated destination fields
  (#124);
• preserved and CI-frozen Prometheus contract with reviewed additive
  current metrics (#126);
• bounded, path-only WAF matched-request logging that omits query values and
  macro-expanded request data (#127);
• HTTP/3 mTLS parity, exclusive ACME challenge selection, compression
  , dependency/CI hardening, and reconciled product documentation.

Artifact verification

• 12 archives: Linux, Windows, and macOS × amd64 and arm64 × lean and full;
• 12 per-archive checksum files plus (25 assets total);
• every GitHub asset digest, aggregate checksum, and per-archive checksum
  independently matched;
• every archive contained its binary, , README, SECURITY, and an
  SPDX 2.3 SBOM;
• all 12 binaries passed SLSA provenance verification and SPDX SBOM attestation
  verification against , source ref
  , and source digest ;
• Linux/amd64 lean and full binaries both reported .

Soak classification

The release gate completed two five-minute smoke scenarios with zero proxy
request errors and bounded UDP-session/goroutine behavior. Per ADR 0005 and the
soak evidence log, this is release-path smoke evidence only; it does not replace
the one-hour/four-hour minimum long-running GA soak evidence.

Known residual work outside the candidate

• response-cache correctness and recertification (#107, #131–#134);
• closed-world lifecycle authority (#89);
• generation-safe dynamic access-log ownership (#98);
• deterministic replacement of scheduler-sensitive Windows deadline tests
  (#185, reopened after an unchanged rerun was needed).

Published for review. A stable requires a fresh
explicit decision, tag, workflow run, and publication review; this RC tag is
never moved or renamed.

Jul.IA v1.30.0

Jul.IA v1.30.0 — Phase 2A Consolidated Soak

🔬 Phase 2A Consolidated 8-Hour Soak Test

The most demanding soak test in the project history: all 10 shipped features exercised simultaneously for 8 hours with zero errors.

Metric | Value -- | -- Duration | 8h0m0s Total requests | 2,120,299 HTTP 2xx | 2,120,299 (100%) HTTP 401 / 403 / 429 / 5xx | 0 Connection errors | 0 Timeouts | 0 Error rate | 0.00% Latency avg | 670 ms Latency p95 | 1,257 ms Latency p99 | 1,560 ms

Features exercised: Proxy, Cache, Rate-Limit, WAF, Auth (Basic), Compression (gzip/Brotli/Zstd), TLS, mTLS, Upstream Health-Checks, OTel Tracing.

🐛 Bug Fixes

• Compression auto-enable: [compression] block with settings but no enabled = true is now automatically enabled (the block implies intent). Fix in parser.go.
• OTel schema-URL fix: Updated semconv import from v1.39.0 to v1.41.0 to match OTel SDK v1.44.0, fixing