docs(#5337): add architecture-review evidence table for #5296 sub-issues

[qqeasonchen]

What this PR does

Closes #5337 (Q1 production validation evidence + Q8 closure criteria for #5296 sub-issues).

This PR adds docs/architecture-review/evidence.md -- a single-source-of-truth evidence table that links every closed sub-issue of #5296 to:

• the PR that landed the change,
• the implementation commit on develop (squash hash),
• the test files (JUnit / TCK) that exercise the change,
• the test command to reproduce the result,
• the CI run that observed the pass/fail (or current status if pending),
• the backend + version under which the test runs,
• the deployment topology (single / multi-instance / crash-recovery),
• the observed result.

Coverage

19 PRs catalogued, of which:

• 16 MERGED to develop (refactor(common): partition eventmesh-common into internal sub-packages (#5298) #5320, feat(arch-guard): enforce architecture rules in FAIL mode + dedicated CI workflow (#5305) #5322, feat(storage): standardize storage SPI capabilities + TCK (issue #5303) #5323, feat(security): unified multi-tenant / security / quota entrypoint (issue #5304) #5324, docs: sync documentation with implementation status (issue #5306) #5326, [ISSUE #5231#5232#5233] Harden ConnectorRuntime against data loss; split connector-api SPI module; add plugin unit tests (#4642) #5328, docs: downgrade Connector Runtime Beta -> Experimental (follow-up to #5326, sync with #5328) #5329, feat(state): control-plane store SPI + deprecate zombie code (#5301 Sub-PR A) #5310, fix(state): persist ReliableDispatcher.pending to DeliveryStateStore (issue #5301 Sub-PR B, fixes #5294 #5295) #5311, feat(state): DeadLetterStore + TaskStore Meta backends (issue #5301 Sub-PR C, fixes #5292 fully, unblocks #5302) #5312, feat(a2a): wire A2A Gateway onto Runtime via TaskStore (issue #5302 Sub-PR D1) #5313, feat(cluster): PARTITION_OWNED_PULL delivery topology (issue #5309, closes) #5317, test(state): cross-store fault-injection E2E for the unified state control plane (issue #5314) #5318, fix(#5295): gate RocketMQ 5 POP broker ACK on distribution completion (PR #5316, with both contributor trailers) #5336, docs(#5341): separate modern CloudEvents HTTP / EventMeshFrame path from legacy TCP / gRPC / OpenMessaging #5343, docs(#5338): wire DeliveryTopology from production config + document control-plane defaults #5344) -- CI on record before the repository-level Actions platform outage
• 3 OPEN-CI-startup_failure (test(#5339): state store durability tests + per-backend failure matrix #5345, feat(#5340): A2A TaskExpirer + MetaAgentCardRegistry + failure-mode test (D2a) #5346, fix(#5342): storage capability matrix + arch-guard plugin boundary rules #5348) -- PR code review passed; CI auto-retries will collect the run once Actions recovers

7 reliability scenarios from #5337 Q1

#	Scenario	Primary evidence
1	RocketMQ 5 POP broker ACK barrier (duplicate ACKs / out-of-order)	#5311 + #5336
2	RocketMQ 4 -- basic delivery ACKs (single + multi)	#5311 + #5317
3	Kafka -- offset persistence (broker-ack / broker-ack-then-crash)	#5311 + #5345
4	Crash recovery (kill mid-delivery, no double / no lost)	#5311 + #5318
5	DLQ failure (exhaust retries -> durable DLQ)	#5312
6	Cursor recovery (restart, per-subscriber cursor)	#5311 + #5345
7	Multi-instance fencing (split-brain, stale partition owner)	#5317 + #5318

All 7 scenarios have at least one MERGED PR with the test implementation.

Items moved to follow-up (per #5337 acceptance)

• [Architecture Review] A2A readiness: TaskExpirer, Meta-backed AgentCard, Runtime E2E, failure testing #5340 D2b (Testcontainers E2E for A2A Gateway) -- tracked at [A2A] D2b: Testcontainers E2E for A2A gateway (issue #5340 follow-up) #5347
• [Architecture Review] Storage capability coverage and architecture guardrail enforcement #5342 Q6 plugin-load-time capability validation -- documented as future work in docs/storage-spi.md (EventMeshSPI loader change needed)
• Scenario 4 on RocketMQ 5.x POP -- Testcontainers-driven restart-then-replay run not yet on record; follow-up implied by [A2A] D2b: Testcontainers E2E for A2A gateway (issue #5340 follow-up) #5347

Current CI status (as of 2026-09-08 09:50 GMT+8)

A repository-level startup_failure is affecting every workflow that requires a GitHub-hosted runner since 2026-09-07 16:23 UTC. 65% of recent workflow runs are startup_failure. Last known successful push to develop was 2b0d7abb1 on 2026-09-03 02:18 UTC. Workflow file content is unchanged between the last successful and current failing runs (verified for architecture-guard.yml). This is a repository-level Actions registry / runner availability issue, not a code-introduced regression.

Acceptance

• An architecture-review/evidence.md (or section in [Architecture Review] Consolidate EventMesh boundaries and production-grade consistency #5296) lists each sub-issue with the 7-column evidence row above.
• Every row has a passing CI run (or documented manual repro) and a backend-version pin.
• Any item lacking evidence is moved to a follow-up issue (not silently closed).
• [Architecture Review] Consolidate EventMesh boundaries and production-grade consistency #5296 closure references this evidence table (planned once test(#5339): state store durability tests + per-backend failure matrix #5345/feat(#5340): A2A TaskExpirer + MetaAgentCardRegistry + failure-mode test (D2a) #5346/fix(#5342): storage capability matrix + arch-guard plugin boundary rules #5348 CI is collected).

Co-authored-by: qqeasonchen qqeasonchen@gmail.com