How can you be a realist and not posit an ontological model? I have a specific technical question about how to formalize models for quantum interpretations.
My question arises from the talk Why I am not a psi-ontologist, by Rob Spekkens at the Perimter Institute, in which he explains in detail the differences between $\psi$-ontic and $\psi$-epistemic models for quantum theory, as described in e.g. this previous question, and, more in depth, in Spekkens and Harrigan's paper Found. Phys. 40, 125 (2010), arXiv:0706.2661.
At the start of the talk (first two or three minutes of the video), Spekkens lays out some of the basic terminology for the formalism, starting with


*

*an operational theory, which 

just gives you a prescription, an algorithm for calculating the probabilities of different outcomes of some measurement given some preparation.

That is, an operational theory just posits the existence of preparation procedures $\sf P$ and measurement procedures $\sf M$, where 


*

*to each preparation procedure $\sf P$ quantum theory associates a density operator $\rho_\sf{P}$, and 

*to each measurement procedure $\sf M$ quantum theory associates a set $\{x^{(\sf M)}\}$ of possible measurement outcomes, and a set of measurement operators $E_x^{(\sf M)}$ which form a POVM,


and the operational theory restricts itself to talking about the probability $P(x^{(\sf M)}|\sf{P,M})$ of outcome $x^{(\sf M)}$ given a preparation procedure $\sf P$ and a measurement $\sf M$, which is described by quantum theory as $$P(x^{(\sf M)}|\sf{P,M})=\rm{Tr}(\rho_\sf{P} E_x^{(\sf M)}).$$
Spekkens then goes on to talk about realist interpretations:

If you favour realism then you might say that's not enough as an interpretation, and we would also like to be able to say that the system that comes out of this device has some physical properties and it's those physical properties that are relevant to the measurement outcome. 
So, I'm going to say that an ontological model of quantum theory is a realist interpretation and it has the following sort of form. It posits ontic states: little lambda is going to denote the physical state of the system that passes between (the preparation and the measurement), and capital lambda is the space of such states.
And, for every preparation procedure you have in the lab, you posit that there is some statistical distribution over physical states of the system that correspond to that preparation procedure.

Then, an audience member (Adrian Kent if I've got it right) intervenes with the following comment:

I would say that to be a realist you don't have to be committed to any ontological model of quantum theory the way you've just defined it. It's a special case of realism.

— to which Spekkens immediately agrees.

I find this deeply confusing. How can you be a realist as regards quantum theory, and not even posit something that can be interpreted as some big space $\Lambda$ that groups the physical states of the system? What description can be more generic or broad than Spekkens' definition of an ontological model? How do those interpretations look like, and where can I find examples of them used in practice in the literature?
 A: I think the realist stance in its broadest sense has to do with getting rid of the observer in the description of "reality". That is, there must be a principle of relativity allowing all observers to share a same reality, even if they do not observe the same thing.
In quantum mechanics the measurement problem makes this incompatible with an ontic physical state associated to a quantum system; one could make the point that saying "quantum state" is already assuming too much onticity so to speak. It seems that a quantum systems gets some spatiotemporal reality only with relation to its measurement, which is where the dependency on the observer appears.
But if we generalize "reality" beyond spacetime, and consider measurements as spatiotemporal anchors of wider non-local elements of reality, then we can be both realist and denying that a system has to be associated to a physical state. It could just be that a system as it is described in QM is a metaphoric representation of something transcending the spatiotemporal language, and that only in measurements do spacetime actually manifests.
This would explain why trying to interpret physically a quantum system implies illogical articulations of spatiotemporal notions (superposition, entanglement).
For an example of research along this direction, see Making sense of a world of clicks (Mohrhoff, 2002).
An excerpt:

Since the beginning of time (in about 1926) it has been argued that QM
is about experience, knowledge, or information,
rather than about a free-standing reality capable of being described
without reference to observers, their information, their interventions
into “the course of Nature”, or their arbitrary decisions as to
where to make the “shifty split” between “system” and “apparatus”.
Why? Because it is such an easy way to establish the consistent
coexistence of extrinsic and intrinsic variables. If the properties of
the quantum world are extrinsic (that is, if they “dangle” from, or
supervene on, something), and if the quantum world is coextensive with
the physical world, then from what can they “dangle”? The obvious
answer: from us, from what we perceive, or from what we know.
For this
easy way out we pay a high price. By safeguarding against empirical
refutation conceptions of space and time that are consistent with
the phenomenal world but inconsistent with the physical world, we make
sure that we won’t discover the spatiotemporal features of the
quantum world. And by rooting the possible value-indicating events, to
which QM assigns probabilities, in the world of sensory experience, we
make sure that we can’t conceive of the quantum world as a strongly
objective, free-standing reality that owes nothing to observers,
information, or our interventions into the course of Nature.

Another excerpt:

In general it can be said that whenever QM requires the addition of
amplitudes, the distinction we make between the corresponding
histories do not exist in the physical world.

A: It's hard to speculate what the questioner might have meant, but certainly they could have meant that a realist does not have to commit to any particular set of ontic states as being the "real ones", they only need to hold that there does exist some sense of "real states."  They could be skeptical that quantum formalism "has it right," and are looking ahead perhaps a thousand years to some totally new theory.  They might even think the "reality" is beyond our ability to comprehend, yet there is "a reality" all the same.  In that way, you could be a realist and still be psi-epistemic.
