An auxiliary HotpotQA boundary probe in the paper's supplement evaluates whether adaptive depth gating offers a distinctive advantage on open-domain multi-hop QA. On the aligned FullWiki-1k slice, Recall@$10$ is $93.4$ for Flat dense, $92.9$ for the Hierarchical baseline, and $94.0$ for Adaptive (heuristic) depth gating. The three systems are statistically close, so the probe finds no distinctive adaptive-depth advantage beyond an already-strong flat dense retriever. The Limitations section reads this as evidence that the curated prerequisite ranking does not transfer cleanly to a denser non-prerequisite graph, so the probe is treated as a boundary condition rather than as a headline result.