Your RAG system is being used for customer support at a software company. Retrieved passages come from technical documentation, community forums, and internal knowledge base articles. Users report that answers about current software behavior are sometimes incorrect because forum posts from 2019 are retrieved with higher similarity than recent documentation updates. What architectural control directly addresses this temporal relevance problem?

B is correct because the temporal relevance problem has two components: retrieval ranking (outdated posts outranking recent docs on similarity alone) and generation awareness (Claude doesn't know which sources are more reliable). The hybrid scoring with recency decay addresses retrieval ranking. Source-type weighting (official docs > forums) applies domain knowledge about reliability. Passing dates and source metadata to Claude enables it to reason about temporal relevance during generation (e.g., 'This forum post from 2019 conflicts with the 2024 documentation; the documentation is likely current'). A is wrong because re-embedding with newer models changes the semantic representation but does not alter the temporal ordering of documents — a 2019 forum post re-embedded with a 2024 model still has the same publication date and can still outrank recent documentation on semantic similarity. C is wrong because a hard one-year cutoff would delete valuable long-term documentation that remains valid — many technical concepts, APIs, and configurations remain stable for years. Age is a factor, not a binary cutoff. D is wrong because chunk overlap affects whether multi-sentence concepts stay together in chunks — it has no relationship to document recency or retrieval ranking.