Hi there! I don’t know if anyone is still listening, but I have made a simpler version of the problem, and found a workaround. At this point I am sure that either the loop over items node isn’t consistent or that the design principle of subworkflows isn’t intuitive.Here is the isolated error workflow:
We can appreciate here that only the last batch is being processed on the second loop, but all items are passing, which means that the node is recieving the items but deciding not to process them.
However, if we encapsulate that loop inside a subworkflow, it does process everything. This is the subworkflow version:
And the subworkflow:
We can see that, in both versions, connections are exactly the same logically, but don’t produce the same result. I am not sure if this is the way subworkflows are meant to work, but in my understanding, I supposed that subworkflow logic should be isomorphic to not encapsulating in a subworkflow at all.
Anyway, this is what I have been able to discover, I hope it helps.