[33/36] Add OC-160 gRPC demo reliability task

[cjohnsto-nz]

Stack PR for the Missio 0.8.0 OpenCollection review queue.

Scope

Adds the docs-only OC-160 planning track for reliable local gRPC demo startup and verification. The task is consistently registered in the goal prompts, progress board, dependency map, gap-analysis index, and dedicated task file.

The acceptance wording permits either 127.0.0.1:50051 or localhost:50051 only when the fixture and collection endpoint agree on the same loopback service. The implementation delivered later in PR 70 binds the fixture to loopback, documents the matching collection endpoint, and exercises all seven packaged gRPC demos through the real execution path.

Review outcome

No source or documentation correction is required on this branch. The reviewer approved the task definition; changing equivalent loopback spelling would add stack churn without changing the requirement or implementation.

Verification

• all OC-160 task, prompt, board, dependency, report, and index references resolve at 1fe1f59
• GitHub build and Socket security checks pass
• PR 67 -> 68 and PR 68 -> 69 three-way stack checks are clean
• GitHub reports PR 68 and its adjacent stack PRs as MERGEABLE / CLEAN

[APKiwi]

Verdict: approve. Internally consistent (task file 16 to OC-160 matches the existing convention, README row resolves, dependency map and board cross-reference), correctly scoped as a demo/verification task, fixtures kept loopback-only. Nit: acceptance says 127.0.0.1:50051 or localhost:50051 while the plan says 127.0.0.1, same thing on loopback, harmless.

[cjohnsto-nz]

Reviewed against the OC-160 task scope and the later implementation in PR 70. No change is required on this branch.

The 127.0.0.1:50051 / localhost:50051 wording is intentional acceptance flexibility: the requirement is that the fixture and collection endpoint reach the same loopback service. PR 70 makes that concrete by binding the fixture to 127.0.0.1, documenting the localhost:50051 collection value, and starting the real fixture on an ephemeral loopback port in integration coverage that drives all seven packaged gRPC demos. That proves the distinction does not leave an unverified endpoint.

The task file, goal prompt, board row, dependency map, and README index are internally consistent. GitHub checks pass, and the PR 67 -> 68 and PR 68 -> 69 stack edges are clean. I have deliberately left the implementation branch unchanged to avoid unnecessary descendant churn.