>>173
Don't know of such process. Any known process either terminates or repeats itself.
Only because you're looking at finite state machines which could technically repeat themselves given enough time (if sometimes much longer than the expected lifetime of the universe).
In my example, the UD is simple, imagine you have a scheduler which each time step creates a new process, the initial program that it runs is defined by index 0, the next process will be the previous number + 1, and so on. Thus at t=0, you have process 0 running, t=1, you have process 0 running for 1 time step, and process 1 being started, at t=2, you have process 0 running for 2 time steps, process 1 running for 1 time step, and process 2 just started, and so on. If you can't think of abstract machines which have unbounded memory (which can grow as it sees fit), just consider that your machine is constantly upgraded with more and more memory. You will obviously say that the machine "terminates" along with the universe, but then you have to define the universe and other things which will require such unbounded concepts which you're trying so hard to avoid.
What if specification had error, and implementation was right? It is common to have outdate comments or mismatched interfaces.
It's possible. It just shows that there was a standard that promises something and a program written by the standard writter which wasn't quite compliant to that standard, even though 'it worked'.