I think the main issue in your model is that you create very complex "families" of tokens. See the example screenshot below showing the parent-child relationships of a token chain. There are child-token of different generation in the same activity which means you probably have loops in the overall cycle. With that many relationships and label access on parents it becomes very tough to follow what repercussions label assignments might have. Assigning a label to some token might also change the value its "great parent's second degree nephew"-token can access, if you understand what I mean.

As it stands, I understand why you are having difficulties to debug the model, but so would anybody else. So my suggestion would be that you rethink how tokens are created. For example, I don't see why you would create a child-token right before a token enters a sink. If the token is no longer needed, the created token should probably be independent (with necessary labels copied over) or even just move the token to other part of Process Flow instead of creating a new one.
