Failed to export span: SDK retry policy missing

Goal

Reduce alert noise when the BatchSpanProcessor logs Failed to export span during brief network hiccups while preserving visibility into real data loss.

Constraints

The SDK currently lacks a retry configuration, so transient failures are logged immediately and can trigger alerts on every occurrence.

Users often rely on the Collector’s retry_on_failure block, but the SDK side remains silent, creating a mismatch in reliability handling.

Unresolved Decision

Whether the SDK should expose a retry or back‑off policy that aligns with the Collector’s capabilities.

Questions

• Should the BatchSpanProcessor provide a public API to configure retry intervals and maximum attempts?
• What impact would adding such a feature have on existing exporter implementations and backward compatibility?
• Which metrics or thresholds would best indicate a sustained export failure that warrants an alert, without flooding ops teams during normal transient outages?