CR 7127924 was filed due to some failures during windows integrations.
The fix for that CR was to run the three @tests in 
   test/tools/javac/diags
in othervm mode to force a file that is intentionally left unclosed to be closed after the test has been run.

The Evaluation in that CR suggests a more sophisticated fix, which could be done for this CR.

Another possible problem with these fixes is that the file that is left open tends to get closed when a gc happens - I presume a finalizer closes it.  It seems like there might be a window between when this file is created by the ProcUnclosedTypeFiles example, and when it is accessed in that example, during which a gc could run and cause the file to be closed.  If this were to happen, it seems like it would cause the current @test to fail.  I tried to cause this this failure but was unable to do so.