Protect critical paths
Keep full-fidelity traces for checkout, auth, and other revenue paths even when sampling elsewhere.
Route rare error codes and high-latency spans to 100% sampling automatically.
Sampling should never hide an outage on the customer journey.
Use dynamic sampling
Increase sampling during incidents and rollouts; scale down when systems are steady.
Feed sampling decisions from Watch.Dog alerts so coverage matches risk.
Validate visibility
Run chaos tests and confirm the signals appear in dashboards and alerts.
Review dropped events regularly to ensure important data is not being thrown away.