End-to-end request-level observability that follows one user action across every microservice it touches.
Distributed tracing instruments each request with a trace ID that propagates across every service the request hits, so the resulting trace is a tree of spans showing the full call graph and timing. A 500ms request that crossed seven services will show which span took the 480ms. OpenTelemetry is the modern standard for emitting traces; popular backends include Jaeger, Tempo, Honeycomb, Datadog APM, and Nova.
In a microservice architecture, no single log file tells the story of a slow request, the slowness lives in the boundary between services. Tracing is the only observability surface that crosses those boundaries cleanly. Teams without tracing diagnose multi-service incidents by guessing; teams with tracing diagnose them by reading the timeline.
See the part of the platform that handles distributed tracing in production.