test(interop): qualify Python and TypeScript reporting artifacts across stable and skew

[bokelley]

Summary

Add the blocking real-process Python/TypeScript stable-and-skew interoperability matrix required before the Reliable Reporting release can merge and publish.

The matrix must qualify immutable installed artifacts, not source checkouts, PR heads, mutable tags, or announcement metadata. The corrected TypeScript implementation from adcontextprotocol/adcp-client#2972 must first be published and identified exactly; the final Python candidate must include accepted rc.4 adoption and #1172.

Required matrix

• Four pinned cells crossing the stable and candidate Python artifacts with the stable and corrected-candidate TypeScript artifacts.
• An installed PostgreSQL-backed Python reference seller exercised over real MCP HTTP by the pinned JavaScript CLI.
• Semantic reconcileReporting() assertions, not snapshot-only or schema-only checks.
• Managed Delivery and Reconciled Billing status/receipt flows, signed webhook retry/replay, account activity, exact revision reads, and deterministic fixture identity.
• A fresh process and isolated database/state per cell; do not share a long-running storyboard context across cells.
• Exact package versions, registry URLs, integrity/hashes, protocol tag/source, runtime versions, commands, results, and retained logs for every cell.
• Stable compatible lanes are blocking. Separately labeled latest canaries remain visible but non-blocking.

Acceptance criteria

• Matrix refuses mutable, unpublished, mismatched, or source-only inputs.
• All four required cells run back-to-back from clean installed environments and pass.
• Failure in any required assertion or cell propagates to one required aggregate check.
• Cross-cell state leakage is prevented and regression-tested.
• Evidence identifies the exact accepted Python and TypeScript artifacts and protocol rc.4 source.
• Matrix is rerun on the final integrated Python main commit and again where required for the release-PR merge commit.

Ordering

Harness preparation and baseline runs may proceed in parallel. Final blocking acceptance waits for the published corrected TypeScript artifact, accepted Python rc.4 replacement, complete #1172, and the reviewed release guard.

Related: #1172, #1193, #1198, adcontextprotocol/adcp-client#2972, adcontextprotocol/adcp#7574.

[bokelley]

Triage

Classification: Test coverage — Python/TypeScript interoperability matrix for Reliable Reporting release
Bucket(s): (cross-repo test infrastructure; no single-bucket label applies)
Status: deferred

What the experts said:

• Pre-classification routes directly to Defer (child-of-open-parent): the issue's own "Ordering" section names feat(reporting): provide an adapter-first ReliableReportingService and conformance kit #1172, accepted Python rc.4 replacement, published corrected TypeScript artifact from fix(reporting): align rc.4 timing and discovery adcp-client#2972, and the reviewed release guard as explicit prerequisites. Expert consultation is not run for issues that gate on unresolved upstreams.

My take: Harness preparation and baseline runs may proceed in parallel per the issue's own guidance, but the four blocking acceptance cells cannot pass until those upstream artifacts are published, immutable, and identified exactly. This resurfaces when #1172 merges.

Blocked-on: #1172 — resurfaces on merge. Also gated on adcontextprotocol/adcp-client#2972 (TypeScript artifact publication) and completion of #1193, #1198 per issue ordering.

---

Triaged by Claude Code. Session: https://claude.ai/code/session_01QraXdCQYjJks7kTmaQNgM2

---

Generated by Claude Code