More interestingly, she claims that the "still-images of actions" are also indexical, because they only show a part of a broader temporal whole action. This is probably the most astute observation of the piece, yet receives relatively little relative attention. This is a key insight, and would be worth expanding on.
First of all, in the semiotics of C.S. Peirce, indexicality is a means through which reference is garnered via causation or indication. For example, an index finger that points to something doesn't mean that thing, it indicates the thing has meaning. The finger is just saying "for the real meaning look over there." Also, if I saw a footprint in the sand, it indexes the person who once walked there, because of the causation stepping there created.
Another aspect to indexicality can happen through part-whole relationships. By showing just a hand, you index the rest of the body (assuming it hasn't been detached....*shudder*).
So, related to Magnussen's point, I'm recalling this particularly salient image in my mind from the book How to Draw the Marvel Way depicting a figure punching in several points throughout the overall action.
What Magnussen seems to imply is that a single snapshot of one part of this event sequence indexes the whole rest of the sequence. I would agree with this in general, though I think it's likely that different places within that progression will be more or less salient as indicating the whole.
For example, in How to Draw..., Lee and Buscema's advice is to use the maximally intense points of that sequence — the ends and beginnings of the action marked "best" or "not bad." These sections of the action seem more representative of the action than the medial parts. In semiotic terms, they would index the overall action better than the parts in the middle, which are less representative of the overall action.
Research seems to have borne out their intuition. Studies have shown that people's comprehension of events is better for the maximally preparatory and completed parts of an action, over that of the middle. In fact, even 10-11 month old infants seem to parse events through these outer boundaries.
Now, it would be unsurprising if Magnussen's statement were attempted to expand beyond the representation in single panels, out to across panels. So, let's say that two panels show both the beginning and end parts of that punch sequence. Here, there is a sense that the whole middle part of the action is indexed by seeing the ends — i.e. you know the middle happened but didn't see it.
This would be, essentially, what McCloud is arguing for with Closure. That, because we know the course of events, we "fill in" knowledge of the whole action by seeing the parts. However, even McCloud acknowledges that not all panel relations are of actions (for example, his Subject and Aspect transitions), though his notion of Closure is extended over all of them.
While I do not believe that "Closure" happens to "fill in the gutter" between panels (for numerous reasons), I do think that part of a represented action might index full actions. I don't know if I'd say that the whole action, including the middle parts, is "manifested" somehow in our minds. However, the reference of a part of an action certainly would index the concept of the whole action, by only being a sliver of it.