I'm just going to close this because it doesn't identify any specific problems and I don't want to build these things to deal with hypothetical problems.

I'm fine with adding more behavior here, but I'd like to drive it by identifying things which are hard to test or hard to understand and then finding the best solution we can for those cases.

The only nitpick I have with tests currently is that this pattern (to test for a throw) is cumbersome:

$caught = null;
try {
  thing();
} catch (Exception $ex) {
  $caught = $ex;
}
$this->assertTrue($caught instanceof Exception);

...but I think the best solution for that involves closures, which we won't have until 2020 or so:

$this->assertThrows(
  'Exception',
  function() {
    thing();
  });

...and that's really just a small quality-of-life improvement.