The monotonic algorithm suggested by the spec is somewhat näive, and has resulted in a lot of confusion and discussion among consumers and implementors. An example here. In fact the spec should probably not recommend an algorithm for monotonicity in the random component at all.

@alizain In the last few months, I've seen multiple instances where developers cited this (flawed) approach for monotonicity as an authoritative source of truth, and made similar mistakes in their own domains: other ULID implementations, ULID adaptations, discussions about future UUID specifications, and more. Please give this PR a review. I think it is important.