Clarify MSTest [DoNotParallelize] deferred execution and cancellation semantics

[Copilot]

[DoNotParallelize] docs described only mutual exclusion, but omitted two scheduler behaviors that affect correctness assumptions and runtime cost: deferred end-of-source execution and early-cancellation skip behavior. This PR updates the conceptual MSTest execution guidance to document those semantics explicitly and scope them correctly to each test source (assembly).

• Behavioral semantics added

  • Documents that MSTest partitions each source into parallelizable and nonparallelizable sets when in-assembly parallelization is enabled.
  • States that the parallelizable set runs first, and DoNotParallelize tests run serially afterward as a deferred tail.

• Execution scope clarified

  • Clarifies that “runs last” is per test source/assembly, not global across a multi-assembly run.

• Cancellation interaction documented

  • Explains that deferred tests might not execute if the run is canceled or aborted before the parallelizable phase completes.

• Existing note tightened

  • Makes explicit that DoNotParallelize is inert when parallelization is not enabled (default behavior).

When you enable in-assembly parallelization, MSTest partitions each test source (assembly) into parallelizable and nonparallelizable tests. MSTest runs the parallelizable set first, then runs the `DoNotParallelize` set one test at a time at the end of that source's run.

Because MSTest runs deferred tests only after the parallelizable phase finishes, a canceled or aborted run can end before MSTest executes deferred tests.

• Fixes MSTest [DoNotParallelize] docs omit defer-to-end ordering and cancellation behaviour #55021

---

Internal previews

File	Preview link
docs/core/testing/unit-testing-mstest-writing-tests-controlling-execution.md	Learn preview

Build report

[Copilot]

Copilot wasn't able to review any files in this pull request.

---

💡 Add a code-review agent skill for context-aware, tailored reviews. Learn more in the docs.