Use merge after parallel work
A workflow can branch so separate steps run along different paths. Place a merge step where those paths should come back together, and connect at least two incoming branches to it. The merge waits for the parallel inputs required by the workflow before the path after it continues.
Use this step to express a control-flow join: independent branches have reached a shared point. It does not automatically combine fields from their outputs into one record, choose the best result, or resolve conflicting values. If the next step needs a particular value, design an explicit step to select or transform the branch outputs.
Check the connections and downstream expectations
If a run remains before the merge, inspect the run history to see whether each expected branch actually started and completed. A branch that failed or never became eligible can keep the join from being reached. Confirm the outgoing connection too, since a successful merge does not run later steps that are disconnected.
After the merge, downstream steps should depend on the fact that the parallel paths have joined, not on an assumed combined payload. Test the workflow with representative branch results and verify exactly what the next step receives. When the goal is to aggregate data, use a supported data operation or an explicit transformation rather than relying on the merge control-flow step.