effect-wide-events

Wide events are single, comprehensive structured records per request that consolidate identity, business context, performance metrics, and outcomes. They replace scattered, low-context log lines so distributed incidents can be diagnosed via queries instead of guesswork.

Design span attributes by capturing identity, user, business, performance, and outcome fields with sufficient dimensionality. Validate that common incident questions, such as identifying which feature caused failures for premium users, remain queryable after instrumentation.

Tail sampling manages observability costs by retaining errors, slow requests, and VIP traffic at 100% while sampling successful, fast traffic. This preserves high-value diagnostic data without retaining every routine event.

Avoid scattered step logs, overly thin events with too few fields, and technical-only telemetry that omits business or user context. These anti-patterns prevent you from correlating distributed incidents and answering real operational questions.

Yes, canonical log lines consolidate one comprehensive event per request across services and dependencies. By capturing high-dimensionality attributes like user and business context, they enable direct querying of cross-service failure correlations.