Observability has become a cornerstone of modern software systems. OpenTelemetry libraries power millions of applications across diverse environments - from IoT devices with limited resources to massive web apps at scale. As libraries that are embedded within and run as part of apps, OpenTelemetry must deliver high performance and maintain API stability, while ensuring efficient resource usage.

In this talk, I'll share my experience from contributing to OpenTelemetry Go. I'll explore how we balance performance, API simplicity and stability while adhering to the OpenTelemetry Specification. Through real-world examples of tackling challenges like minimizing heap allocations to reduce the GC overhead, I’ll share the strategies, principles, and development practices that make this possible.

These lessons extend beyond OpenTelemetry, providing actionable guidance for anyone building high-performance Go packages.