Preflight summary counter miscounts (prints FAIL=1 with zero failing checks)

tools/preflight_us_release_gates.py printed Overall: AT_RISK (exit 2) [FAIL=1, AT-RISK=2, clean=0] on a run whose own --json-out records exactly: selection_carryover PASS, zero_support_preview PASS, export_mass_parity_risk AT_RISK, smoke_probe_support PASS — no FAIL, one AT-RISK, three passes. Exit code and JSON are right; the human-readable tally is wrong. Cosmetic but it reads as a failing gate to an operator. Repro inputs: v9.2 feed 61b115c0…, populace main 4746aba.

Start in tools/preflight_us_release_gates.py and reproduce the reported human-readable summary using the cited v9.2 feed and populace main inputs. Trace how the displayed FAIL and AT-RISK counts are computed, compare them with the --json-out records, and confirm done when the summary reports zero FAIL, one AT-RISK, and three PASS checks while preserving the correct exit code and JSON.