Siddhant Khisty

In a real production networking incident, a platform team receives over 200 alerts in 10 minutes across gateways, routes, retries, and upstream services. Every alert is technically accurate, and yet O11Y just amplified the noise.

This talk draws on multiple real-world production incidents to explore recurring observability anti-patterns that surface as alerts triggered by incorrect thresholds, missing correlation, high-cardinality telemetry that overwhelms storage, and dashboards that are difficult to navigate

We will examine why these failures surface at system entry points like network gateways and network routes, where distributed complexity converges. Attendees will leave with practical design patterns for building observability systems that accelerate root cause identification, reduce ambiguity during incidents, and support human reasoning rather than multiplying symptoms at scale.