Cardinality spikes in observability systems, particularly in Prometheus, refer to sudden increases in the number of unique metric series, which can significantly impact system performance and resource usage. Cardinality, defined as the number of elements in a set, becomes problematic when a metric with low or medium cardinality suddenly shifts to high cardinality, leading to increased data generation and potential system overloads. This can cause memory errors, system crashes, and higher operational costs, as observed with billing in Grafana Cloud Metrics based on the number of active series. Such spikes often occur when new metrics are instrumented with excessive contextual labels like "user_id," creating numerous series for a single metric. Effective cardinality management is crucial for maintaining system efficiency and controlling expenses. Grafana Labs provides resources and tools to help users understand and manage cardinality, offering dashboards and configurations for monitoring and optimizing metric usage.