On reading the text in H.225.0 describing the construction of a globally unique identifier, it seems to me that duplicates are possible if the clock has not advanced since the last identifier was generated.
This may occur if the clock ticks at a slow rate (say ten milliseconds), if the CPU is sufficiently fast, or there is more than one CPU.
In the case of a slow clock, say ten milliseconds, this can be compensated for with the low order bits of the time field being cleared and used as a sequence number for identifiers generated in the same clock tick.
In what has been called a "Carrier Class Gateway," with tens of thousands of SCN ports and more than one CPU, it may be possible to generate a duplicate identifier, for a CallIdentifier, within the same 100-nanosecond interval.
In this event, a possible action is to increment the clock sequence if the time "has not advanced" rather than "has gone backward." Thus, if the current UTC is less that *or equal* to the saved value, or the saved value was lost, the clock sequence is incremented.
Douglas Clowes OzEmail Interline