I looked around to see if I could find a duplicate of this question and I found this Probability of Gambler's Ruin with Unequal Gain/Loss
which is very close but lacking in the main generalization aspect I had in mind.
Suppose a gambler has initial fortune $M > 0$, and win probability $p$ for each round with gain $a > 0$ if the round is won and loss $b > 0$ otherwise, where $M,a,b$ can be any positive reals without any restrictions such $a,b,M \in {\mathbb N}$ which was assumed in the linked question and the answer and paper that treats it.
As usual, the gambler is ruined (and the game stops) the first round $t$ when their current fortune $X_t$ becomes $\leq 0$.
The probability of eventual ruin should still be $1$ if and only if the expected winnings of each round is non-positive. But other than that, I wonder if there can be any non-trivial progress to express/compute probability of eventual ruin (e.g. perhaps similar to the paper in the linked question, but in any event better/more insightful than naïve series where each variable $X_t$ is treated individually) as a function of general $M,a,b$. The main obstacle is the chaotic nature of the sign of $M + wa - (t -w)b$ when $w \leq t$ are positive integers but $M,a,b$ might not even be linearly dependent over the rationals.