This has started to fail for GTK since we switched to use adwaita media controls, but I think this was passing by coincidence. The test itself is wrong and that's probably the reason why it fails in the other platforms too. The test expects that the changes that happen after removing the controls attribute from video element are sync, but they aren't because a mutation observer is used by media controls script. Then it assumes that panel display property is set to none after controls attribute is removed, but we are actually removing the media controls tree from the DOM in that case so the panel is null at that point. This worked in GTK before because all those checks that were supposed to happen after removing the controls were done before the handleControlsChange() was called.