docs(planning): keep test gate decisions single-sourced

Make OQ-3 the sole decision point for Redis 8.8.1 gate layering while OQ-10 consumes that result as an input to the broader test strategy.

Constraint: Limit this follow-up to the current CodeRabbit finding on PR #394.

Rejected: Repeat the gate split in OQ-10 | Two writable descriptions can drift into conflicting planning decisions.

Confidence: high

Scope-risk: narrow

Tested: git diff --cached --check; OQ section structure and single-source assertions; LF line endings.

Not-tested: No Rust build or runtime test was run because this follow-up changes one planning Markdown line; new remote CI is pending after push.

Co-authored-by: OmX <omx@oh-my-codex.dev>

Author: AlexStocks

.planning/OPEN_QUESTIONS.md

@@ -55,7 +55,7 @@
 ## OQ-10 [OPEN] 统一测试策略的落地取舍
 - **背景**：#325 提案 `kiwi-test-harness` + Oracle 对比 + property-based + 确定性网络模拟（turmoil）。
 - **待确认**：
-  - Redis 8.8.1 exact Oracle 对比如何在 fast gate 与完整定期门禁之间分层（见 OQ-3）；
+  - Redis 8.8.1 Oracle 的 gate 分层与兼容矩阵以 OQ-3 为唯一决策入口，本项只消费该结论；
   - 是否值得 port Redis 官方 TCL suite，还是以 `resp-compatibility` 为 fast gate；
   - Python 测试保留为多语言 client 验证，还是逐步迁到 Rust harness；
   - 集群/哨兵功能当前可用程度。