Here's how ash processes accelerators:

  // The flow of accelerators in ash is:
  // . wm::AcceleratorFilter() sees events first as it's a pre-target handler.
  // . AcceleratorFilter forwards to its delegate, which indirectly is
  //   implemented by AcceleratorRouter.
  // . AcceleratorRouter may early out, if not it calls through to
  //   AcceleratorController. This stop may stop propagation entirely.
  // . If focus is on a Widget, then NativeWidgetAura gets the key event, calls
  //   to Widget::OnKeyEvent(), which calls to FocusManager::OnKeyEvent(), which
  //   calls to AshFocusManagerFactory::Delegate::ProcessAccelerator() finally
  //   ending up in AcceleratorController::Process().
  // . OTOH if focus is on a content, then
  //   RenderWidgetHostViewAura::OnKeyEvent() is called and may end up consuming
  //   the event.
 
In the mus world the closes we can get to this is to register accelerators for both pre and post processing. The pre should call to the Router, the post to the AcceleratorController.