Answer Matchers Determine What Test-Time Sampling Appears to Do

Test-time scaling is often framed as a policy problem. The system chooses how many samples to draw, when to stop, and whether to rerank, assuming that the measured outcome reflects the policy. On open-ended factual tasks, however, the measured outcome often reflects the answer matcher instead. We use the exact decomposition $R_S=C_S+A_S$ to split plurality error into non-coverage and aggregation failure. A matcher false negative enters both terms with opposite $S$-dependence, creating an apparent conflict between candidate acquisition and selection. On Llama-3.1-8B/TriviaQA, one normalization step affects $46\%$ of all generations. Correcting it moves single-sample accuracy $0.185\to 0.645$, turns total error from an increase ($0.781\to 0.869$) into a decrease ($0.354\to 0.225$), and reverses the measured gain of adaptive stopping from $-5.50$ to $+1.53$ points. The distortion is predictable rather than idiosyncratic. For the exposed class of matching rules, it equals the fraction of a model's correct answers in the affected style times its accuracy. This relation has unit slope over 11 cells, replicates on EVOUNA's held-out human labels, and separates contemporaneous 8B open-weight models by an order of magnitude. Plain exact match is exposed in $15$ of $33$ cells and SQuAD-family normalization in none. We release the canary that detects the failure, which standard parsing metrics miss, together with two further cases in which a cheap scalar accounts for most of an effect attributed to an expensive mechanism. We compare process reward models with trace length and learned trajectory observers with prefix likelihood.