At scale, observability breaks when SLIs and metrics mix different behaviors and lose clear meaning.
Complexity grows: more event types, extra labels, and rising cardinality. That bloats queries, slows evaluation pipelines, and distorts Prometheus, PromQL, and Elastic metrics.
Why this matters: Teams must treat metrics like paid resources. Constrain index scopes. Curb label cardinality. Preserve SLI semantics.
