I wonder if for things like the issue with not having an oracle database to test against warrants a formal rule in unit testing.
"Don't assume that all integration tests will be run in a stable environment" Should our tests be more reflective of the environment or should they be ridged so that all tests environments are the same.
I think maybe instead of having pass, failed, or skipped as test outcomes, maybe another outcome could be "UnSupportable". This would indicate that this test was skipped because the underlying features meant to be tested cannot be tested.
Our unit tests then should check to see if Active directory is available, or if an oracle instance is nearby.