Assuming you mean something like the attached, where two nodes (the purple ones) are directly connected to the same output of a preceding node (the trigger), then the payload will indeed be cloned and each branch gets its own copy. Changes to the payload in one branch will not affect the payload in the other branch, nor is there a way to merge the changes together.
When one node’s execution depends on the output of a node before it, the answer is almost always to run the nodes serially (one after the other). That is of course less performant than parallel execution. But if you can describe your use case, I may be able to recommend a better approach.