OpenTelemetry after the hype: Making observability a platform primitive

Mesut discusses how OpenTelemetry usually enters a platform as a good idea but slowly becomes hard to reason about as more teams start using it. The talk is based on real platform engineering work and focuses on questions that appear after the initial setup is done. It explores what breaks first, where observability becomes noisy instead of helpful, and why metrics, logs, and traces behave differently in practice. The session also examines the OTel Collector as something that quietly turns into a platform boundary. The talk shows why ownership and clear limits matter more than adding instrumentation.

Key takeaways:
• Why OpenTelemetry issues usually show up late
• What changed in the ecosystem leading into 2026
• Where observability design causes friction in Kubernetes platforms
• How to simplify observability without losing signal