In https://bugs.openjdk.org/browse/JDK-8286345 we have a timing out / failing test. The test itself doesn't appear stuck - it's the child process that gets stuck, which causes the test itself to timeout.

A solution could be to try and kill any descendants before deciding to timeout the test, using a pre timeout alarm. If that doesn't get the test unstuck, then we should proceed as normal and timeout the test.

Additional benefit of un-stucking the test is that the err/out streams of stuck child get read and printed out in the parent process and we can see the output that otherwise goes missing.