Alert on impact and risk
A useful alert explains what failed, who is affected and what the first check should be.
Group the same incident
Repeated checks should update one event instead of filling inboxes and hiding new problems.
Tune with real history
Review false alarms, missed incidents and response time. Thresholds should change as the system and workload change.
Official technical sources
These sources support the technical explanation. The right fix still depends on the environment where the problem appears.