Close the live search evidence loop before tuning retrieval

Minimal production usefulness for internal semantic search requires one closed evidence loop: repeated live queries must succeed, newly created or updated sources must become searchable within a bounded freshness target, and a small independently authored question set must retrieve a directly supporting passage that can be opened at its source. Fix the first failing gate before comparing embedding models, tuning rerankers, or expanding benchmark machinery. This is a minimum functional bar, not a claim about privacy, large-scale performance, or eventual search architecture.