Summary

We should consider targeting a release soon to deliver several valuable improvements that have been completed or are near completion.

Proposed Improvements

🚀 Performance Enhancement

• #1794: Improve histogram/summary performance under contention - Significant throughput improvements for high-contention scenarios by striping observationCount across multiple instances. Benchmarks show prometheusClassic improving from ~5,126 ops/s to ~14,624 ops/s (~185% improvement) and prometheusNative from ~3,854 ops/s to ~7,406 ops/s (~92% improvement).

📊 Observability

• #1738: Add GC duration histogram - Introduces jvm_gc_duration_seconds metric to track JVM garbage collection pause durations. Uses OpenTelemetry semantic convention-aligned buckets (0.01, 0.1, 1, 10) with labels for gc name, action, and cause, enabling detailed monitoring of GC behavior and latency spikes.

⚙️ Configuration

• #1828: Support environment variables in properties - Adds environment variable support for configuration (e.g., IO_PROMETHEUS_EXPORTER_HTTP_SERVER_PORT) while maintaining full backward compatibility. Also introduces snake_case property naming (recommended) while keeping camelCase working.

🔧 Flexibility

• #1800: Allow metrics with the same name (Option Sample Values Should be Double, Not Float #2) - Enables registration of metrics with the same name, with validation at both registration and scrape time.

Benefits

These improvements collectively enhance:

• Performance under high-load scenarios
• Observability of JVM runtime behavior
• Configuration flexibility for different deployment environments
• API flexibility for metric registration

cc @jaydeluca @jack-berg