Verdict from the flags

the level is set by the most severe flag that fired: identity_failed, wiped_out, checks_waived give 'not_usable'; breached, sample_too_small, no_out_of_sample, edge_inside_cost_band, lookahead_differences, in_out_gap, implausible_return, resolution_divergence give 'weak'; stopless_share_high, concentrated_one_trade, concentrated_one_period, concentrated_one_day, concentrated_one_year, carry_dependent, search_budget_exceeded, final_slice_not_held, performance_decaying, rule_profile_stale give 'use_with_care'; nothing fired is 'no_warnings'. The first three codes, worst first and in catalogue order inside a severity, are the reasons