Decide whether a cluster incident is still diagnosable, and abstain when it is not. Use when investigating a node that already rebooted, an intermittent failure with no pattern, or any incident where logs have rotated and the evidence window has closed. Also use to sanity-check a confident diagnosis.
