It turns out that one cannot have two fundamentally different parallel set membership relations $\in$ and $\in^*$, if they both satisfy ZF with respect to the common language, for in this case they must in fact be isomorphic. 
In particular, in this situation, they cannot satisfy different
theories.

**Theorem.** If $\in$ and $\in^*$ are membership relations each
satisfying the ZF axioms in the combined language, then $\langle
V,\in\rangle$ is isomorphic to $\langle V,\in^*\rangle$.

**Proof.** One can see that $\in^*$ must be well-founded, since for
every set $x$ the $\in^*$-class $\{y\mid y\in x\}$ has an $\in^*$-minimal
element, by the $\in^*$-foundation axiom, and such an element is an $\in^*$-minimal element of $x$. Similarly, $\in$ is
well-founded with respect to $\in^*$.

If one also knows that the relations $\in^*$ and $\in$ are set-like
with respect to each other, then we may consider the Mostowski
collapse $$\pi(x)=\{\pi(y)\mid y\in^* x\},$$ which is an
isomorphism of $\langle V,\in^*\rangle$ with some transitive class
$\langle M,\in\rangle$. But in fact, we must have $M=V$ since if
every element of a set $z$ is $\pi(y)$ for some $y$, then let $x$
be the set with $\in^*$-elements $y$ whenever $\pi(y)\in z$. This set
exists by replacement, using that $\in$ is set-like with respect to
$\in^*$, and it follows that $\pi(x)=z$. So by $\in$-induction,
every set is in $M$, and we have the desired isomorphism.

But for the general case, one doesn't know at first that the
relations are set-like and so more care is needed. Consider the situation of
$$\pi:\langle V_\alpha,\in\rangle\cong\langle
V^*_{\alpha^*},\in^*\rangle,$$ where $\alpha$ is an $\in$-ordinal
and $\alpha^*$ is an $\in^*$-ordinal and $V_\alpha$ and
$V^*_{\alpha^*}$ are the rank-initial segments of the universe as
constructed with respect to the two membership relations. Since
transitive sets are rigid, these isomorphisms are unique when they
exist. We can always extend to one more level by considering the power
sets, which map across by their pointwise action. And we can take
unions at limit stages. It cannot be that one side runs out of
ordinals before the other, for in this case we would have a
bijection of the whole universe to a set, either in the
$\in$-universe or in the $\in^*$ universe. So this provides the desired isomorphism, as well as a proof that the relations are in fact set-like. $\Box$

Since isomorphic models have the same theory, it follows that:

**Corollary.** If $\in$ and $\in^*$ are membership relations
satisfying the ZF axioms in the combined language, then $\langle
V,\in\rangle$ has the same theory as $\langle V,\in^*\rangle$.

I think this theorem is classically known. Similar ideas are used
in my paper:

<cite authors="Hamkins, Joel David; Kikuchi, Makoto">_Hamkins, Joel
David; Kikuchi, Makoto_, [**Set-theoretic
mereology**](http://jdh.hamkins.org/set-theoretic-mereology/), Log.
Log. Philos. 25, No. 3, 285-308 (2016).
[DOI:10.12775/LLP.2016.007](http://dx.doi.org/10.12775/LLP.2016.007),
[ZBL1369.03047](https://zbmath.org/?q=an:1369.03047).</cite>