4.2. Pick the metric that matches the question🔗

Available metrics are:

total_ns             self_ns
mean_ns              median_ns
p95_ns               max_ns
total_heartbeats     self_heartbeats
alloc_bytes          peak_live_bytes

Use mean, median, or p95 for per-call latency when the call count still represents the same workload. Use total or self time when the amount of recorded work is itself the target.

Heartbeat metrics focus on Lean-managed work. Allocation metrics are meaningful only when both captures use hooks with the same allocator semantics.