Your team runs a high-traffic ASP.NET Core API instrumented with Application Insights. Finance flags that ingestion costs have spiked, and you want to reduce telemetry volume. However, the on-call engineers report that during incident investigations, related request, dependency, and exception records for the same operation are sometimes missing, making distributed traces incomplete. You need to cut ingestion volume while preserving statistically accurate metrics and keeping correlated telemetry items together for a given operation. Which approach best meets these requirements?