[Ruiqi Wang (RachelWanggg)]

Motivation & Context

A fan-out target selector that returns the same index more than once gets that target run once per entry. FanOutEdgeRunner.ChaseEdgeAsync maps the indices to sink IDs and keeps every entry, so the target appears twice in the DeliveryMapping and MapInto enqueues the envelope twice into its queue. A selector returning [0, 0] runs target 0 twice on the same message. Routing by tags hits this whenever two tags map to the same executor. With an agent executor as the target, the message appears twice in the agent's prompt, and the agent runs twice if the TurnToken is routed the same way.

Description & Review Guide

• What are the major changes? FanOutEdgeRunner.ChaseEdgeAsync drops repeated target IDs with Distinct(StringComparer.Ordinal) once the IDs are resolved, from the selector or from the target list when there is no selector, and before the envelope-target and CanHandle checks. Distinct keeps the first occurrence, so delivery order follows the order in which each target is first named. This is the change Python: Deduplicate fan-out selection results so each target runs once #8739 made to the Python FanOutEdgeRunner, which also covers both paths.
• What is the impact of these changes? Each selected target receives the message once. An edge without a selector whose target list names an executor twice now delivers to it once too. Target lists and selections without repeats deliver exactly as before. AddSwitch is unaffected, since SwitchBuilder already collects its case indices in a HashSet<int>. No public API change.
• What do you want reviewers to focus on? Whether repeated targets should be deduplicated, as here and in Python, or rejected. Rejecting them would turn workflows that run today into errors, so this PR deduplicates.

Three tests in EdgeRunnerTests.cs, all failing on main:

• Test_FanOutEdgeRunner_DeduplicatesAssignedTargetsAsync: a selector returning [1, 0, 1] over ["executor2", "executor3"] delivers to executor3, then executor2, once each. On main the deliveries are ["executor3", "executor2", "executor3"].
• Test_FanOutEdgeRunner_DeduplicatesRepeatedSinksAsync: no selector, target list ["executor2", "executor2"], one delivery. On main there are two.
• Test_FanOutEdge_TargetAssignedTwiceRunsOnceAsync: a workflow built with AddFanOutEdge<string> and a selector returning [0, 0] runs the target executor once. On main it runs twice.

Local verification on macOS arm64, .NET SDK 10.0.402, dotnet test --project tests/<project> -f net10.0, branch on top of deecbcb. The main column is this branch with FanOutEdgeRunner.cs reverted to main.

Unit test project (Microsoft.Agents.AI.<name>.UnitTests)	main	This PR
Workflows	858 passed, 3 failed (the new tests)	861 passed
Workflows.Declarative	1115 passed	1115 passed
Workflows.Declarative.Mcp	not run	90 passed
Workflows.Generators	not run	36 passed
Hosting	not run	181 passed
Hosting.A2A	not run	154 passed
Hosting.AGUI.AspNetCore	not run	12 passed
Hosting.OpenAI	not run	585 passed
Hosting.AzureStorage	not run	2 passed, 12 skipped (Azurite not available)
Foundry.Hosting	not run	520 passed, 1 skipped (existing Skip)
DevUI	not run	21 passed, 2 skipped (existing Skip)
CosmosNoSql	not run	25 passed, 54 skipped (Cosmos DB Emulator not available)

• Mutation checks, EdgeRunnerTests only: removing the Distinct fails all three new tests. Applying it only when there is no selector fails the two selector tests. Applying it only to selector results fails Test_FanOutEdgeRunner_DeduplicatesRepeatedSinksAsync. Rebuilding the list in target-list order, or sorting it, fails Test_FanOutEdgeRunner_DeduplicatesAssignedTargetsAsync.
• dotnet format <csproj> --verify-no-changes passes for Microsoft.Agents.AI.Workflows and Microsoft.Agents.AI.Workflows.UnitTests.
• dotnet build src/Microsoft.Agents.AI.Workflows --warnaserror passes for net10.0, net9.0, net8.0, netstandard2.0 and net472 with 0 warnings.
• Not run: net472 tests (Windows only), integration tests, and unit test projects that do not depend on Microsoft.Agents.AI.Workflows.

Related Issue

Fixes #9431

AI Assistance

• No material AI assistance was used.
• This is an AI-assisted contribution. I reviewed, understood, and verified all submitted content and accept responsibility for it.

Contribution Checklist

• The code builds clean without any errors or warnings
• All unit tests pass, and I have added new tests where possible
• The PR follows the Contribution Guidelines
• This PR links to an agreed issue with no competing open PR, or the Related Issue section documents a trivial-change or repository-automation exception.
• This is not a breaking change. If it is a breaking change, add the breaking change label (or add "[BREAKING]" to the title prefix, before or after any language prefix). A workflow keeps the label and title prefix in sync automatically.

[Vincent Biret (baywet)]

Thanks for the contribution!

[Copilot]

🟢 Approval recommended

The localized fix covers both routing paths with focused regression tests and no identified blocking issues.

0 open findings

🧠 Review effort: Balanced

[Vincent Biret (baywet)]

depends on #9455 for merge