Improve crashloop detection

When workloads repeatedly fail there should be improve crashloop detection that surfaces actionable signals (frequency, recent failures, affected tasks) so users can more quickly identify and remediate failing workloads.