All snaps in a pipeline are initialized in parallel. After initialization, each snap waits for input from its upstream snap. As data becomes available, the data is processed by the snap and written to the downstream snap. The control flow is driven by the availability of data to process. Snaps connected to each other will process a document in the same order as the order they are connected to each other in.
If there are multiple unconnected branches in the pipeline, each of them processes data in parallel. There is no synchronization between pipeline branches. Disconnecting two snaps which are next to each other functionally makes them unconnected branches.
