question

Because you are sending the items out through random ports (by percentage), many items are send to racks that don't actually have a slot to store them in (matching labels requirement). As a result a lot of the items don't actually get pulled and remain in the queues. Eventually the queues end up being totally filled with these items. As no new items can enter the queues the model comes to a stop. You can also see this in the sources that lead directly into the racks. Many items will stay in the sources, as they couldn't be assigned a slot in the rack they were send to by the round robin.

If you want to keep the distribution of items between the racks the same, you could either write some custom code in the form of

if(item.Type < 25)
{
    // Percentage Code for the respective ports
}
else
{
    // Percentage Code for the respective ports
}

The code inside the brackets could be copied from the current "by percentage" option, with the percentages scaled to 100% in total for each bracket.

Alternatively, you could push the items to a list in the "Send to Port" option. You could vary the partition by assigning a label based on the current percentage distribution in the entry trigger and use this to determine the list partition. Then each rack(-pair) would only pull from the respective partition (i.e. one partition for turkey1+turkey2 and so on).

A third option might be to not send the items directly to the rack. Instead have a queue for each pair of racks, from which the correct rack can then pull the item (first available).