I used long tracks here for initialization to try and keep the period low, which is good for instructions but bad for cost. Doing a cost style initialization where one arm does all the work could be better for cost and area, with the downside of multiplying the instructions on all the steady state arms. A small benefit of increasing the period would be that the elemental part of the machine could operate at pseudo-period 3 while the metal part can run at pseudo-period 4.