Unstable runs of Test Cases

I think it would be considered bad practice to do 20 applications in a single test case. I wouldnt even test a single application entirely in a single test case.

A test case should be seen as a unit test, and you are doing maybe 100 units (assuming 5 units per 20 applications) in a test case.

Why are you trying to do so much in one place instead of separating them?