[SergeyMenshykh]

Motivation

Fix the intermittent .NET Framework failure in Factory_LoopbackInvocations_ReuseHandlerAndDisposeSeparateSessionsBeforeReturningAsync. During cleanup, the local test server can attempt to accept another connection after its listener has stopped, throwing InvalidOperationException: Not listening.

Description

• Handle InvalidOperationException only around AcceptTcpClientAsync() and only when shutdown is requested.
• Keep unrelated request-processing failures visible instead of suppressing them during shutdown.
• Add a deterministic regression test that verifies request-processing errors still propagate during shutdown.

No production code changes. The .NET Framework suite passed; .NET 10 validation was interrupted.

AI assistance: implementation, regression test, and PR description.

Related Issue

N/A - trivial change: correct shutdown handling in a test helper without changing production behavior.

Observed in the failing CI job for #10244. This PR fixes the test cleanup race, not the Foundry scaffolding feature.

[Copilot]

🟢 Approval recommended

No active review issues block the test-only shutdown-race fix.

No open issues 🎉

What changed in this PR

Fixes a shutdown race in the .NET loopback MCP test server without changing production code.

Changes:

• Handles InvalidOperationException from AcceptTcpClientAsync() only when shutdown is requested.
• Adds a deterministic regression test confirming that request-processing failures still propagate during shutdown.

File	Description
dotnet/​tests/​Microsoft.Agents.AI.Workflows.Declarative.Mcp.UnitTests/​DefaultMcpToolHandlerLifetimeTests.cs	Adds targeted listener-shutdown handling and a regression test for error propagation.

_{Review effort: Balanced}

[github-actions]

MAF Automated Review — Iteration 1

Result: No findings
Scope: full PR (1 commit(s)): 37f16d057036
Model: gpt-5.6-sol

Overview

The PR narrowly handles the loopback test server's accept-after-shutdown race without changing production code. The cancellation filter and catch placement constrain suppression to AcceptTcpClientAsync(), while disposal continues awaiting the request task and the new synchronized regression test confirms request-processing failures propagate unchanged. No publishable Critical, High, or Medium issue was established.

Reviewed the supplied pull-request change set across correctness, security/reliability, architecture, and failure behavior.
No publishable findings remained after source verification for this scope.