A system can have a balance between positive and negative feeback. If it has a mix of both, there's amplification, not necessarily a runaway. (The balance between solar input and radiation to space, among other things provides negative feedback)

Moreover, it isn't even just a multiplication problem. There are different styles of feedback - proportional, integral, differential - and those latter two can come with different time scales

It's obvious that pushing the same direction for a hundred years can be much bigger a deal than pushing a hundred times as hard in the same direction for a day, but it's also true of a hundred-times-as-strong push lasting for, say, two years. Or, depending on the different feedbacks, the hundred times as hard for a day could have a bigger effect.

And all of that is without going nonlinear!