When NCrunch 2.19.0.4 on distributed nodes, or running more than one thread locally, I am seeing many spurious failures with the message:
"This test was not executed during a planned execution run. Ensure your test project is stable and does not contain issues in initialisation/teardown fixtures."

This appears to only have started happening after a recent update from NUnit 2.6.3 to NUnit 3.2.0.

Many of these tests have no init/teardown, and appear stable and stateless.

Running these tests "manually" (i.e. right clicking and "Run selected task in a new task runner process") makes it go green.

If I disable my connection to all distributed nodes, and force all tests to run locally on one thread, this problem appears to go away.

Thanks for the quick reply.

Using 2.20 appears to allow me to run multiple threads locally, but it fails in the same way on distributed nodes - so it is an improvement over having to run a single thread locally.

I've tried various combinations for CopyReferencedAssembliesToWorkspace / PreloadReferencedAssemblies / etc but they don't seem to help. Reducing the number of processing threads down to 1 on the grid node does not help either.

The only discernable pattern is that it's roughly the same subset of tests (~400/6000) every time, but I cannot see what could cause it from the code in these tests compared to other tests in the same project.

I'm not using any NUnit3 specific functionality - the test code in question did not change, unless there some setting in NCrunch itself around NUnit3 (I did not find any)?

I've been running disconnected from the grid node using 6 threads locally for around two hours now, with no spurious failures - I'll continue doing so for a while to try to confirm it's a grid-specific issue.

Yes, updated grid node to 2.20; I also tried isolating to one grid server and restarting it in case there was some environmental issue causing it, but after restart it is still showing the same issues.

I'll re-enable grid processing later today and submit a bug report with it.

There does not appear to be any common pattern to the failing tests - they are a mixture of Test, TestCase and TestCaseSource; some in classes with [SetUp], some with [OneTimeSetUp], and some without any initialisation.

Apologies for the delay in replying back here - it would appear that installing .Net 4.6 and the visual studio 2015 build tools on the ncrunch grid nodes has fixed the problem!

Will continue to monitor and will update if this changes, but for now it's looking good.