docs(protocol): pin the queue-full cap finding to the served-config evidence

[Mearman]

Summary

• The queue-full verification row said the effective cap was "dynamically raised, value unknown". The Statsig evaluations cache contains no tengu_harbor_kite_limits entry — the override has never been served to this account — so the effective cap is the code default of 50, and the 55-queued non-firing is queue accounting (hold-parked messages don't count toward the undelivered-peer-message queue), not a raised cap.

Test plan

• Cache checked: no harbor_kite key in either evaluations file
• CI verify passes

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
🔒 Security Review	✅ Completed 2026-09-13T04:08:56.105044Z	8110d08	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.