@birtles can you take a look at the changes here? The bit I'm not sure about it:

If we create a ready promise, we need to
wait for it to resolve. Otherwise, an error is potentially triggered
(race condition) during tear down of the test as it will cancel a
pending animation, which in turn, rejects the ready promise. Any
unexpected promise reject triggers a harness error.

I believe I managed to create a minimal reproduction, which appears to reject during cleanup in Firefox Nightly:

// Modified css/css-animations/CSSAnimation-ready.tentative.html so this was the only test in it:
promise_test(async t => {
  const div = document.createElement('div');
  document.body.appendChild(div);
  t.add_cleanup(() => {
    div.remove();
  });

  div.style.animation = 'abc 100s';
  const animation = div.getAnimations()[0];
  const readyPromise = animation.ready;

  // Finish without waiting for ready.
}, 'Testing');

@birtles FYI Kevin realized that Chrome was missing the step to mark the promise as handled, hence the unexpected top-level rejections. So he has reverted the WPT test changes in the downstream Chromium CL and this PR is no longer valid (I will close it once the downstream change lands).

Upstream change landed with no changes to WPT, closing this.