Top 10 Observability Mistakes

1. This tweet warns that observability systems break down when teams skip request-level tracing, use unstructured logs, omit deployment/service metadata, and overuse high-cardinality labels.
2. It calls out sampling that hides rare failures, traces without timings, missing SLOs, and metric/alerting strategies that react to symptoms instead of bottlenecks.
3. The close is to stop treating logs, metrics, and traces as separate systems and stitch them together with shared identifiers.

Relevance

High-signal infrastructure advice for Norviq and Loci/backend work: if you want reliable incident response and postmortems, this reinforces hard requirements around request IDs, structured logging, and shared tracing metadata.

Notes

Engagement is currently low but non-zero: 2 likes, 1 repost, ~200 views.