The token-cap diagnostic is the paper's first-use protocol for comparing two retrieval systems after applying a shared token-aware serialization policy to their returned contexts. Both systems serialize their retrieved items under the same rules and the same token budget, so any remaining difference can be attributed to what was retrieved rather than to how it was packed into the prompt. The diagnostic is analyzed separately from the strict-parity headline comparisons, which is what lets the paper keep traversal effects distinct from token-budget effects.