Stationary many subsets of $\kappa^+$ whose order type is a cardinal and whose intersection with $\kappa$ is an inaccessible cardinal Is anything known about the consistency strength of the following statement?


*

*$\kappa$ is a Mahlo cardinal and there is a stationary set of $a \in \mathcal{P}_\kappa(\kappa^+)$ such that $a \cap \kappa$ is an inaccessible cardinal and the order type of $a$ is $(a \cap \kappa)^+$?


This statement follows from $\kappa^+$-supercompactness of $\kappa$ and also from subcompactness of $\kappa$.  It is a strengthening of the principle $\text{Pr}(\kappa^+)$ that was shown by Burke in "Generic embeddings and the failure of box" to imply $\neg \square_\kappa$, so a lower bound on the consistency strength is the existence of two Mahlo cardinals.
 A: Well, here is a very slight weakening of your $\kappa^+$-supercompactness upper bound, to the assumption merely that $\kappa$ is  nearly $\kappa^+$-supercompact. This hypothesis is strictly weaker than $\kappa^+$-supercompactness, but still, under this assumption, the set of such $a$ is stationary as you desire.
Specifically, a cardinal $\kappa$ is nearly
$\kappa^+$-supercompact, if for every transitive $M$ of size
$\kappa^+$, closed under $\lt\kappa$ sequences, there is an elementary embedding $j:M\to N$ into a
transitive set $N$, with critical point $\kappa$, such that
$j''\kappa^+\in N$. It follows that $j''\kappa^+$ has your
property with respect to $j(\kappa)$ in $N$, and so the set of
$a\in P_\kappa(\kappa^+)$ with your property will have measure one
with respect to the induced $M$-measure. It follows by the usual
normality argument that the set of such $a$ is stationary, as you
desire, since any club set $C$ (or the function defining it) can be placed into such an $M$ and so
$j''\kappa^+\in j(C)$, meaning that $j$ of the set of $a$ meets
$j(C)$, and so there is such an $a$ in $C$.
The nearly $\theta$-supercompact cardinals were introduced by
Jason
Schanker in his dissertation (written under my supervision), along with the
weakly measurable cardinals,
and Jason had noted that the near $\theta$-supercompactness
hypothesis suffice in many arguments that previously had used full
$\theta$-supercompactness. The present case is an additional
instance of this phenomenon, where near
$\kappa^+$-supercompactness suffices in place of
$\kappa^+$-supercompactness.
Meanwhile, I am unsure exactly how the nearly
$\kappa^+$-supercompact cardinals relate to the subcompact
cardinals, and in any case perhaps a more dramatic weakening is possible.
It may be interesting to note that a recent result proved jointly by Jason Schanker, Brent Cody, Moti Gitik and myself shows that it is relatively consistent with ZFC that the least weakly compact cardinal $\kappa$ is also nearly $\kappa^+$-supercompact (but $2^\kappa\gt\kappa^+$), which might suggest something about the nature of these cardinals. 
