[ric-oliv]

The MSBuild job runs msbuild.exe without -m, so it builds the whole Windows solution filter on a single node. The dotnet CLI passes -maxcpucount on its own, which is why the .NET legs already use every core.

From the job's binlog on #5793:

• msbuild ran 21.4 minutes on 1 node: Restore 8.2, Build 12.4 and Pack 0.3.
• 5.8 of the Build minutes are 175 dotnet workload list calls from OnlyTargetInstalledWorkloads in Directory.Build.targets, one per project and target framework, all waiting on each other.
• The .NET (win-x64) leg builds the same filter with dotnet build in 3 to 4.5 minutes.

#4300 added this job to catch build breaks that only happen in Visual Studio, like RuntimeInformation.RuntimeIdentifier failing under .NET Framework MSBuild. With -m it's still the same MSBuild building the same projects, and Visual Studio builds in parallel too.

The job took 22.8 to 29.5 minutes on the last 6 runs, so with the other open CI PRs it would be the next job to hold up the build workflow.

-m also turns on node reuse, which leaves MSBuild's worker processes running after the step. Nothing later in the job uses them, so the command passes -nr:false too.

Measured on the first run, against 3 runs from the same evening without -m:

	Without -m	This PR
MSBuild job	29.6 to 30.3 min	26.0
Run MSBuild step	23.1 to 23.6	18.4
Nodes in the binlog	1	4
Build target	12.4	8.3
Restore target	8.2	8.8

• The Build target got 4 minutes faster, not the 8 I expected. Its last 2.5 minutes are Sentry.Samples.Maui for Android on a single node, which ci: build the app samples in their own job #5783 moves out of this solution filter.
• Restore didn't change, because NuGet still extracts the packages into the user profile on C:. ci: install .NET and NuGet packages on a faster drive on Windows #5782 now moves them next to the workspace for this job too, and on its own run the step took 14.7 minutes.
• With both, I expect the job to take about 12 to 13 minutes.
• The build output is the same as a run without -m from the same evening: 44 warnings with the same codes, no errors, and the same 49 packages.

#skip-changelog

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 76.13%. Comparing base (b41b96c) to head (04cf32e).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #5796   +/-   ##
=======================================
  Coverage   76.13%   76.13%           
=======================================
  Files         528      528           
  Lines       19579    19579           
  Branches     3829     3829           
=======================================
+ Hits        14906    14907    +1     
+ Misses       3791     3790    -1     
  Partials      882      882           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.