The verifier exits nonzero on four conditions: a missing expected channel, a recording that is too short, a disallowed sample rate, or missing required events. Each is a hard stop because downstream analysis would be meaningless without them.
Everything else is a report, not a verdict
Channel names, sample rate, duration, per-channel RMS and peak-to-peak, and detected markers are all reported for a human to read. The verifier makes no modeling or clinical claims — it only tells you whether the recording is structurally sound.