Declare subset ranking when trading candidate breadth for speed

For a bounded lexical search, reducing candidate verification can materially improve latency, but sorting the surviving candidates does not produce global relevance or global recency. Return the candidate budget and an overflow signal, label rankings as within the nominated subset, and teach clients to refine person, time, session or phrase filters when overflow occurs. Measure both latency and answer-bearing evidence on independently selected questions before choosing the budget. This is an approximate retrieval trade-off, not a recall estimate; session diversification after truncation cannot recover useful sessions that never entered the subset.

Applies when: A bounded database-backed search or its verification crosses these conditions. Keywords: candidate budget subset relevance recency latency recall lexical