This is a weird way to represent friction, and the description does not do a good job of explaining it.

Friction, or drag, must always have a sign opposite to velocity ( otherwise, it'd be speed - velocity is speed in a direction ), and must always be smaller than it, physically. If friction can somehow be bigger than negative velocity ( impact with a heavier object? ), (a) it's no longer friction and (b) yes, it can reverse velocity.

If it's supposed to model cumulative friction over time, the model also breaks down - friction at any moment can still never be bigger than negative velocity, so can never sum to more than negative velocity.

Or maybe you just need a different model to explain what you want us to calculate, because this one doesn't work.

Backstories are nice, but they should be somewhat believable.

velocity will always be one number, negative or positive, as will friction.

Friction can be negative?

expected 1.8000000000000003 to deeply equal 1.8

The test cases should allow for floating point inaccuracies instead of using strict comparisons, see the docs here.