A few thoughts:
- Signifiers required (location-agnostic):
In my experience, Accordions are usually used when you want to offer a large amount of textual information, but reduce the amount of visual processing required to sift through it. FAQs are a common usage example. Sequences (eg. step-by-step processes) are another.
To me, that suggests that to find a good design, you should focus on how you can make it easy to manage a larger, rather than a smaller set of Accordion Items, because that's how people will typically use it. So then you need a design that can signify as much of the overall sequence of items as possible. Showing 3 at a time, in an Accordion with 12 items, is probably not a good experience. I think people will want to be able to select any item from the sequence at random, to be able to adjust design settings for it.
By extension, the affordance of "Rearrange ordering" should also be considered for usability at scale. I tend to think that it's less common to want to move an item all the way from #1 to #12 in the sequence than it is to move it 2 or 3 steps. And of course, drag & drop is a convenient metaphor for doing so. For many people, especially those with a trackpad as their primary touch interface, dragging + scrolling a list in one action is often a frustrating experience, as compared with dragging but not needing to scroll the list to find the position you want to drop in. If you use Spaces in OS X, think about whether you prefer to drag a window to the next space by holding it at the screen edge, or whether you more often find yourself using Mission Control to achieve the "window on a different space" result. I think that's a really comparable (and frustrating) UX to dragging + scrolling, because both diminish the user's sense of control over the UI through an unpredictable interaction speed once they hit the edge of the visible UI (the speed is not the same as the speed they used to drag the object, so it's a jarring effect).
And I think it might be sufficient to provide the reordering functionality through a separate interaction. Maybe, through some kind of simple list control that just allows for reordering, presented in a modal context. That would give you more vertical space to show more or all of the sequence, too, as you wouldn't need the extra height for your button padding (inspect, duplicate, delete aren't really common parallel actions when you want to reorder items).
On a separate track:
As I was thinking about the "see the whole sequence at once" signifier, I started thinking about how mature design tools handle this sort of thing. For a long time, Adobe tools have allowed the user to dynamically resize the height of panels that show a list of items (eg. layers). So you might like to prototype a height-adjustable design panel? If someone wanted to see 12 accordion items at once, they could just drag a handle on the horizontal top border of your design surface, to make the whole panel taller. They could then do all the work they need to do (eg. reorder), then resize it down again. Like you can in the Apex text editor, for instance Harder to do on smaller screens, but that's true for most design tools anyway. You might find that relaxing this particular UI constraint opens up a lot of possibilities for you. Just an idea