The issue is that resmokelib.errors.ServerFailure is considered fatal and converted into a StopExecution exception.

Proposal is to raise a TestFailure instead of a ServerFailure when the mongod is found to exit with a nonzero code when being restarted.

# Raise this after calling setup in case --continueOnFailure was specified.
if not teardown_success:
    raise errors.ServerFailure("%s did not exit cleanly" % (self.fixture))

The effect of this is that when LeakSanitizer causes the mongod to exit with code 23, we will mark the test that triggered the leak as having failed, but continue running the remaining tests.