Observation frequency, evidence weight and authority answer different questions. (A) Conceptual human, agent and robot streams distinguish missing outcomes, repeated source records and dependent telemetry. Time is unscaled. (B) One retained TP-01 synthetic participant contains 40 rows and 20 known episode IDs; each pair of diamonds groups into one circular ID mark. This is known-ID grouping, not a general proof of independence. (C) A separate numerical illustration uses Beta(34,6). Its lower fifth percentile is approximately 0.74915, below 0.75; equivalently F(0.75) exceeds 0.05, so the exact declared gate returns REVIEW. The panels do not constitute one executed clinical pipeline. Neither a statistical estimate nor a passing numerical gate supplies clinical sufficiency or permission to act.
