A checkout request calls five services. Logs show errors in several services, but the team needs the causal path and time spent in each call. Which telemetry is most useful?

1. An hourly CPU utilization chart
2. Distributed traces with propagated trace context
3. Uncorrelated logs sorted by local clock time
4. A counter of errors grouped only by service