First, define a new time coordinate $\eta$ through integrating $$ d\eta = \frac{c dt}{a(t)}.$$ This time coordinate is often called conformal time. With this coordinate, the condition for a radial, null trajectory is $$ \frac{dX}{d\eta} = \sqrt{1-K X^2}.$$ So far I haven't done anything new, just simplified my life by defining a new time coordinate. But this is a differential equation: we can solve it the physicists way of rearranging the differentials as $$ \int \frac{dX}{\sqrt{1-K X^2}} = \int d\eta = \eta$$ where have chosen the boundary conditions $(X_0,\eta_0)=(0,0)$ to fix the integration constants. But the LHS is solvable! In particular, it says that $$\eta = \begin{cases}
\sin^{-1}(X) & K =1 \\
X & K = 0 \\
\sinh^{-1}(X) & K = 1
\end{cases}$$
Equivalently, inverting these expressions,
$$X = \begin{cases}
\sin(\eta) & K = 1 \\
\eta & K = 0 \\
\sinh(\eta) & K = 1
\end{cases}$$
So we can see the interpretation of your observation. When $K=1$, the X coordinate has a maximum value, and is more like an angle than a true radial coordinate. That's why the velocity was imaginary for large enough $X$: the manifold just doesn't exist there! In fact, changing coordinates $X=\sin(\varphi)$, the spatial part of the metric you wrote is exactly the metric of a three sphere when you restore the $d\Omega$. That's not a coincidence, and deeply related to the fact that $K=1$ is the case of de Sitter spacetimes. Check out the Wikipedia page on de Sitter space for more information.