[eli-r-ph]

💡 Motivation and Context

On v1, the $trace_id and $span_id that the SDK takes from the active OpenTelemetry span are merged into the call's own properties, so they win over the same keys in context tags and super_properties. Across the v1 SDKs, a value the caller sets (on the event, in a context or globally) wins over a value the SDK adds. posthog-node and posthog-go already let caller and context trace IDs win over the active span's, and posthog-rs is moving to the same rule.

• The span IDs now fill with the other values the SDK sets: after the call's properties, context tags and super_properties, before before_send.
• The SDK adds both IDs or neither: if any layer sets $trace_id or $span_id, it adds neither, so an event never pairs one trace's ID with another span.
• capture, capture_ai and capture_exception on Client, and AsyncPosthog, follow the rule. Client.capture_exception still always adds the span IDs.
• The 8.0 migration guide and a changeset describe the change from 7.x (the span beat context tags there).

💚 How did you test it?

• New wire tests (sync and async) with a real active OpenTelemetry span, over capture, capture_ai and capture_exception: both IDs added when no layer sets them; a call property, a context tag or a super property wins and stops both derived IDs; before_send sees the derived IDs and can change them. The caller, context and super cases fail without this change.
• pytest posthog/test (5326 passed), ruff check, ruff format --check, mypy and make public_api_check pass locally.

📝 Checklist

• I reviewed the submitted code.
• I added tests to verify the changes.
• I updated the docs if needed.
• No breaking change or entry added to the changelog.

If releasing new changes

• Ran sampo add to generate a changeset file

🤖 Agent context

Autonomy: Human-driven (agent-assisted)

Authored with Claude Code (Claude Opus 5.5), directed by @eli-r-ph.

🤖 Generated with Claude Code