Suppose we have a commutative square $\require{AMScd}$ \begin{CD} A @>{f}>> B\\ @V{h}VV @V{k}VV \\ C @>{g}>> D \end{CD} of quasi-categories, such that $f,g,h,k$ are cofibrations in the Joyal model structure (i.e. these are monomorphisms of simplicial sets). Suppose further that $A,B,C,D$ have all finite coproducts, and that $f,g,h,k$ preserve these finite coproducts. Let $E$ be the homotopy pushout of $f$ along $h$ in the Joyal model structure. Assume that $E$ is a quasi-category. Does $E$ have all finite coproducts, and do the functors $B\to E, C\to E, E\to D$ preserve these?
I think the answer should be yes. A model for $E$ could be the following. We know that $B\cup C \subset D$ is the ordinary pushout, hence a homotopy pushout, since the Joyal model structure is left proper. Then we can take $E$ to be the smallest sub-simplicial set of $D$ such that all squares \begin{CD} \Lambda^k[n] @>{}>> E\\ @V{}VV @V{}VV \\ \Delta[n] @>{}>> D \end{CD} for $0<k<n$ have a diagonal filler $\Delta[n] \to E$. I think an $n$-simplex $\sigma:\Delta[n] \to D$ is then in $E$ if and only if there is some surjection $\delta:[m] \to [n]$ such that all edges of the spine of $\delta^*\sigma:\Delta[m] \to D$ are in $B\cup C$.
We have an initial object $0\in A$ which is also initial in $B,C$ and $D$. This should give the initial object of $E$, but I have a hard time proving that $E(0,X)$ is contractible, even for $X \in B \setminus C$. It seems to get quite messy to show this by hand, and it would be nice if there is either a slick argument or a reference I could use.
Edit Simon's example convinced me the answer to the question is probably no, and that at least my construction of $E$ is wrong as stated. I'd like to add the condition $B \cap C = A$, to more closely reflect the situation I am interested in.