An at least pretty good solution to computing parallel curves of Béziers is given in [1]. This can give very accurate results with a modest number of segments, using techniques that are fast to evaluate and not terribly complicated. I know it's possible to do better in terms of number of segments, but it's not obvious to me how to make that faster than doing curve fitting and error bounding.
Thanks for the reference! The current solutions are a bit scattered around the internet. I wonder if it’s worth making a collection and comparison of approaches
[1]: https://raphlinus.github.io/curves/2021/02/19/parallel-curve...