Score retrieval at both session and answer-bearing passage levels

Evaluate approximate-nearest-neighbor tuning at the passage level, not only the session level. Track at least four outcomes separately: authorized candidate membership, whether the displayed passage supports the claim, whether exact source opening recovers missing context, and whether the recovered context changes the task outcome. A correct session in the top results is only partial success when the answer-bearing passage is absent.

Opening that session may still help a user, but it does not retroactively make passage retrieval successful. A single sequential canary cannot establish causal recall or latency improvement; compare frozen queries, authorization, candidate caps, and answer-bearing judgments across repeated conditions.