Worked on the longhorn/longhorn-manager repository to address alert fatigue for Kubernetes operators by refining the event handling around snapshot cleanup. Focused on backend development using Go, the contribution involved normalizing event types so that intentional snapshot cleanup actions no longer generated misleading Warning events. This change ensured that only unexpected failures would trigger Warning alerts, while routine, expected cleanups would be marked as Normal, reducing unnecessary notifications during recurring backup operations. The approach aligned with existing event downgrade patterns in the codebase, preserving critical visibility for real issues while improving the overall operator experience and system observability in production environments.