Let $(M,g)$ be a 4-dimensional Riemannian manifold which is Kähler with respect to two independent complex structures $J_1, J_2$. Why if these induce the same complex orientations then $(M,g)$ is hyperkähler?
The only reference I found for this fact is the mathoverflow question https://mathoverflow.net/a/436867/508697, but the answer is quite obscure to me. In particular I cannot see neither why having two independent parallel and self-dual $(1,1)$-forms forces the 3-bundle of self-dual forms to be flat, nor why this fact implies the existence of an hyperkähler structure on $M$.
The MO answer uses linear algebraic facts in dimension $4$ and considerations about parallel sections of some bundle.
Lets unpack them, the calculations for the linear algebra part are kind of annoying so I will just give the statements. Let $(M,g)$ be an oriented Riemannian manifold of dimension $4$. For $x\in M$ let $*_x:\Lambda^k(T_xM)\to \Lambda^{4-k}(T_xM)$ denote the Hodge star induced by the orientation. Then:
So if $J_1, J_2$ are two complex units on $T_xM$ let $\omega_1,\omega_2$ be the associated $2$-forms. If $J_1, J_2$ induce the same orientation you can check $*_x(\omega_i)=\omega_i$ for this orientation. By Gramm-Schmidt there is some linear combination that makes the $2$-forms orthogonal to each other and norm $1$. Since linear combinations preserve self-duality you get anti-symmetric $I_1, I_2$ with $(I_i)^2=-\Bbb1$ and $I_1I_2+I_2I_1=0$. This implies that $I_1, I_2, I_1I_2$ are $3$ complex units satisfying the same relations as $i,j,k$.
Doing this at every point gives you an almost-hyperkähler structure. Why are the units parallel? The MO answer argues that this follows from dimensional reasons (if an $n$ dimensional euclidean vector bundle has an $n-1$ dimensional flat subbundle then the bundle itself is flat). Perhaps it is simpler to argue as follows:
Then from $J_1, J_2$ being parallel you get that $I_1, I_2, I_1I_2$ are parallel and the structure is actually hyperkähler.