I have been doing CTW parallelization improvements, and noticed that some of the AWT clinits run and initialize graphics stack. This is awkward for a few reasons:

 1. We might be running on headless environment and these clinits could fail, shrinking the CTW testing scope
 2. There are dependencies in graphics stack initializations that break the stack -- in one case, I have seen the VM crash due to uninitialized AWT lock, because randomized CTW runner managed to execute clinints in unusual order.

I think we should be running CTW tests in AWT headless mode to begin with.