To begin with I don't think the mathematics of the time would support the kind of statistical analysis that we are applying, but I do think magi would have the general sense of how likely botches were, and most especially how dangerous, and that both of these would scale with the power of the ritual involved.
I think we also need to step back here a moment and acknowledge that the game designers and authors are also fallible, and that people wo are thinking about ways to tell an interesting story for a game may not always catch the long term statistical implications of their mechanisms. Personally until something else comes forth I plan to use my house rule above, where being involved in the exact same ritual repeatedly (same target for the same basic effect every time) will make botches less likely- it does not change the storytelling dynamics but patches the long term implications. I do think I will allow the same special circumstances expertise to apply to essentially the same spell at differing magnitudes as well.