Each process should have a way to identify if tests are being run Process separation targets resilience for errors. After implementing the resilience, it should be tested. The test likely involves invoking the error. The error should be likely only invokable from testing. This means there should be a flag for question "are tests being run?" that each process can access. Example: WebGL needs to test cases for GPU process hang and GPU process crash. These error conditions should be invocable from the tests, but not otherwise.
<rdar://problem/75166219>