Having to split every validation test output in 3 is a bit unfortunate but this works for me.

. . . Hmm, actually, can we set the mock terminal in the tests to not emit this code consistently except for a specific test for this behavior? Up to you whether that seems worth it.

It depends on test intention actually. If the intention is to test real life e2e scenario, and Mac output is currently indeed different, it's better to keep them split as is.

Unless we "unseal" Terminal implementation, and override only diff members, or make it configurable, the mock implementation can easily deviate and bug will be missed. Given that most of us is developing on Win, Mac and Linux issues can be unnoticed.

From my experience, it's worth to keep truly e2e scenarios unmocked. For specific tests, for example DisplayNodesShowsCurrent, DisplayNodesOverwritesWithNewTargetFramework it might be reasonable to use mock, however i don't have a full picture to make a decision for them.

Unless there is more feedback on open items, this PR is ready to be merged.