Error Rate Percentage vs. Absolute Error Count for Worker Alerts

When configuring monitoring for Cloudflare Workers, the goal is to identify critical failures while minimizing notification noise. The platform allows for different alerting triggers based on request outcomes, but the choice of metric significantly impacts the signal-to-noise ratio.

Using a percentage-based threshold smooths out transient spikes during high-traffic periods, preventing alerts for a small number of failures that do not represent a systemic issue. However, this approach may mask critical failures during low-traffic windows where a 100% error rate on a few requests remains below the percentage trigger.

Conversely, absolute error counts capture every failure regardless of traffic volume but can lead to alert fatigue during traffic surges where a baseline error rate is expected.

• Which metric provides better visibility for low-volume, high-criticality endpoints?
• How does the behavior of CPU time limit exceptions influence the choice between these two alerting strategies?