Here's a model with 2 pullcontrols ( mpull for intTwice, mpull2 for Singleproc2)
Plus m2 to provide the individual proctime for IntTwice and back stations.
As you can see, this kind of matrialflow with backloops causes a deadlock
Precisely, it is necessary to write for the SingleProc that he would not pass the part until the item passes the second time through IntTwice. + reset the value after passing.
I think a buffer in the backloop could faciliate the tendensy of deadlocking.
Nevertheless locking/ using intTwice + back as backloop for every Mu will considerably reduce
the line performance as you are doubling the cycletime by using these 2 stations twice for ever MU !