For sleep sort, the time taken is a constant function of the size of the list; it only depends on the value of the maximum element of the list. T = max(list) + smaller terms associated with scheduling timeouts. So, it's O(max(list)). Although, the "smaller terms" I'm ignoring blow up when the size of the list exceeds your computer's resources. I'd guess that spaghetti-sort is O(sum(list)).
If it is scheduled by the kernel then the complexity is hidden in the scheduling algorithm.
> The O(1) scheduler was used in Linux releases 2.6.0 thru 2.6.22 (2003-2007), at which point it was superseded by the Completely Fair Scheduler.
https://en.wikipedia.org/wiki/Completely_Fair_Scheduler
> The fair queuing CFS scheduler has a scheduling complexity of O(log N), where N is the number of tasks in the runqueue. Choosing a task can be done in constant time, but reinserting a task after it has run requires O(log N) operations, because the runqueue is implemented as a red-black tree.
[0] https://rosettacode.org/wiki/Sorting_algorithms/Sleep_sort