Looking for self-similar, radial solutions to a certain nonlinear wave equation in $\mathbb{R}^{d+1}$ leads to studying the following linear ODE in the self-similar variable $a\in[0,1)$:
$$(a^2-1)Q''(a) + \left(3a-\frac{d-1}{a} \right)Q'(a) + \frac{3}{4}Q(a)=0$$
*Mathematica* tells me that the solution set is spanned by 
$$\operatorname{span}\left\{ {}_2F_1(1/4, 3/4, d/2, a^2),a^{2-d}i^{2-d}{}_2F_1(5/4-d/2, 7/4-d/2, 2-d/2, a^2)\right\},$$
where ${}_2F_1$ is the Gauss hypergeometric function. 
I have two questions:

1. The second hypergeometric function is undefined for even $d\ge 4$ as then $2-d/2$ is a nonpositive integer. What happens then to the set of solutions?

2. For odd $d\ge 3$, by checking each case individually on *Mathematica*, I want to conjecture that the hypergeometric solutions simplify so that the set of solutions is given by 
$$\sum_{\pm}c_{\pm}a^{2-d}(1\pm a)^{d/2-1}P^d_{\pm}(a),$$
where $P^d_\pm(a)$ is a polynomial that is non-vanishing at $a\in\{-1, 1\}$ of degree $d/2-3/2\in \mathbb{N}_0$. Is this conjecture true or false?