This is follow up to a question answered here https://math.stackexchange.com/a/4224037/168758
Fix $\lambda \in [0, 1)$. Define $m_{-1}(\lambda) = m_0(\lambda) = 1$, and for $k \ge 1$, define $m_k(\lambda) \in \mathbb R$ by
$$ m_{k}(\lambda) = \frac{1}{k(1-\lambda)^2} \Big((2k-3)(1+\lambda) m_{k-1}(\lambda) - (k-3)m_{k-2}(\lambda)\Big). \quad (\star) $$
It is easy to deduce that $m_1(\lambda)=\dfrac{1}{(1-\lambda)^2}(-(1+\lambda)+2) = \dfrac{1}{1-\lambda}$ and $m_2(\lambda) = \dfrac{1}{2(1-\lambda)^2} = \dfrac{1}{2(1-\lambda)^2}((1+\lambda)(1-\lambda)^{-1}+1)=\dfrac{1}{(1-\lambda)^3}$.
Question. What is a general formula for $m_k(\lambda)$, perhaps in terms of well-known (special ?) functions, valid for all $k \ge 1$ ?
Computing several iterations gives \begin{align} m_2&=\frac{1}{(1-\lambda)^3}\\ m_3&=\frac{1+\lambda}{(1-\lambda)^5}\\ m_4&=\frac{1+3\lambda+\lambda^2}{(1-\lambda)^7}\\ m_5&=\frac{1+6\lambda+6\lambda^2+\lambda^3}{(1-\lambda)^9}\\ m_6&=\frac{1+10\lambda+20\lambda^2+10\lambda^3+\lambda^4}{(1-\lambda)^{11}}\\ m_7&=\frac{1+15\lambda+50\lambda^2+50\lambda^3+15\lambda^4+\lambda^5}{(1-\lambda)^{13}} \end{align} and so on. The denominator increases by power of 2 each iteration, and the coefficients of the numerator follow a nice Pascal-triangle like pattern, so I [looked it up on the OEIS ][1] and it turns out they are the Narayana numbers, [which follow the formula ][2]
\begin{align} \mathrm N(n, k)=\frac{1}{n} {n\choose k}{n\choose k-1}. \end{align} It is not hard (translation: I don’t have a legitimate proof) to then arrive at the result \begin{align} m_k(\lambda)=\frac{1}{(1-\lambda)^{2k-1}}\sum_{i=0}^{k-2}\mathrm N(k-1, i+1)\lambda^i, \quad k\geq 2. \end{align}
Hopefully this was without error, let me know if you find one. [1]: https://oeis.org/search?q=1%2C%201%2C%201%2C%201%2C%203%2C%201%2C%201%2C%206%2C%206%2C%201&language=english&go=Search [2]: https://en.m.wikipedia.org/wiki/Narayana_number