If, in the extremely unlikely event that, you manage to generate more than 2^80 ULIDs within the same millisecond, or cause the random component to overflow with less, the generation will fail.

Given the width of the random space is 80 bits, the latter is guaranteed to happen first (or at the same time).

There is also no guarantee that the number randomly generated is not very close to 2^80 -- should implementations consider a maximum number they will accept from the random number generator, and that they will change random numbers greater than to? (Or choose a number lower than)?