Python: fix(workflows): preserve all trace contexts in FanInEdgeRunner aggregation

[badhope (weed33834)]

Problem

FanInEdgeRunner collected trace contexts and source span IDs using the singular backward-compat properties (msg.trace_context / msg.source_span_id), which return only the first element of the plural lists:

# Before (buggy)
trace_contexts = [msg.trace_context for msg in messages_to_send if msg.trace_context]
source_span_ids = [msg.source_span_id for msg in messages_to_send if msg.source_span_id]

When a message arriving at a fan-in already carries multiple trace contexts (e.g. from a prior fan-in aggregation in a nested topology), all but the first context per message are silently dropped. This breaks distributed tracing span links for any workflow with nested fan-in groups.

Fix

Iterate over the plural fields (trace_contexts / source_span_ids) and extend the aggregated lists so every trace context and source span ID from every source message is preserved:

# After (fixed)
trace_contexts: list[dict[str, str]] = []
source_span_ids: list[str] = []
for msg in messages_to_send:
    if msg.trace_contexts:
        trace_contexts.extend(msg.trace_contexts)
    if msg.source_span_ids:
        source_span_ids.extend(msg.source_span_ids)

Test

Added test_fan_in_preserves_multiple_trace_contexts_per_message which:

1. Sends a message carrying two trace contexts through a fan-in edge group (simulating a prior fan-in aggregation)
2. Sends a second message with one trace context
3. Asserts the target executor receives all 3 trace contexts and 3 source span IDs (not just 2)

This test fails on the old code (assert 2 == 3) and passes with the fix.

All existing workflow and edge tests continue to pass (55 passed; the only 2 failures in the broader suite are pre-existing and caused by a missing openai optional dependency, unrelated to this change).

Checklist

• Code follows existing style conventions
• Unit test added and passing
• No breaking changes — the singular properties still exist for backward compatibility; only the fan-in aggregation path now uses the plural fields
• No new dependencies introduced

[Copilot]

Pull request overview

This PR fixes distributed tracing propagation in the Python workflow engine’s fan-in edge runner by ensuring nested fan-in aggregations preserve all upstream trace contexts and source span IDs, rather than silently dropping everything except the first element per message.

Changes:

• Update FanInEdgeRunner aggregation to iterate over WorkflowMessage.trace_contexts / source_span_ids and aggregate all entries.
• Add a regression test that simulates nested fan-in behavior by sending a message containing multiple trace contexts and asserting the target receives the full aggregated set.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
python/packages/core/agent_framework/_workflows/_edge_runner.py	Adjusts fan-in aggregation to preserve multiple trace contexts/source span IDs per incoming message.
python/packages/core/tests/workflow/test_edge.py	Adds a regression test to ensure fan-in preserves multiple trace contexts per message.

[badhope (weed33834)]

Addressed in 6fdc8b5. Trace contexts and source span IDs are now paired per-message via zip(msg_contexts, msg_span_ids, strict=False) before flattening, so a message with mismatched counts only drops its own orphans instead of shifting all subsequent pairs out of alignment.

[badhope (weed33834)]

Addressed in 6fdc8b5. Removed the execute() override entirely. The handler now captures trace data from ctx._trace_contexts and ctx._source_span_ids on the WorkflowContext passed to it, avoiding the fragile override of Executor.execute().

[badhope (weed33834)]

@microsoft-github-policy-service agree