effect-wide-events

Wide events (canonical log lines) are single comprehensive records per request containing all relevant context, replacing many scattered log lines. They optimize for structured querying instead of text searching, letting one query answer questions like checkout failures for premium users on a new feature flow.

Use Effect.annotateCurrentSpan or Effect.annotateLogs to attach context to the active span. Include identity, user, business, performance, and outcome fields; extra log-message arguments go into the message body and are not auto-indexed as queryable dimensions.

No. OpenTelemetry is a protocol and SDK for standardizing telemetry collection and delivery; it does not decide what gets logged or add business context automatically. You must explicitly annotate spans with user, business, and outcome data.

Tail sampling retains telemetry selectively after a request completes: keep 100% of errors, requests slower than p99, and VIP users, while sampling only 1-5% of fast successful requests. This preserves debugging value while controlling observability costs.

Logs optimized for writing scatter context across many low-dimensionality lines, forcing grep-style searches across services. Wide events fix this by attaching all context to one high-dimensionality event per request, enabling structured queries.