A process statement does not "create new processes after termination".
If your belief that it is doing this is based on print statements,
read Stephan Merz's recent post to the "Pluscal Print" thread.
Otherwise, I suggest that you write a very simple multiprocess
algorithm in which "every process is supposed to run one time and then
terminate" that "keeps creating new processes", and then study its
TLA+ translation until you understand what the algorithm is actually
doing.