POC of managing a graph where the number of sources is varying, and sources can move from graph to graph

Consider the following state:

flowchart TD
    Entity1 --> Source1
    Entity1 --> Source2
    Entity2 --> Source3

We want to be able to move Source2 from Entity1 to Entity2, like so:

flowchart TD
    Entity1 --> Source1
    Entity2 --> Source2
    Entity2 --> Source3

To achieve that, we'll use the following structure of actors:

flowchart
    subgraph Pauseable Source 1
        Source1 --> Valve1 -- KillSwitch1 --> BroadcastHub1
    end

    subgraph Pauseable Source 2
        Source2 --> Valve2 -- KillSwitch2 --> BroadcastHub2
    end

    subgraph Pauseable Source 3
        Source3 --> Valve3 -- KillSwitch3 --> BroadcastHub3
    end
    
    BroadcastHub1 -- BCH_KillSwitch1 --> MergeHub
    BroadcastHub2 -- BCH_KillSwitch2 --> MergeHub
    BroadcastHub3 -- BCH_KillSwitch3 --> MergeHub

    MergeHub --> ProcessorStage

    ProcessorStage --> OutputSink[Output Sink]

• Source: is the actual source of data. Can be Source.range(), an AmpqSource.committableSource, or otherwise
• Valve: allows easy pause-and-resume on sources, for example, when you move the source from one graph to another, so that you'd be able to resume from the same place when you open the valve again.
• [KillSwitch]: this is for when you actually want to kill the origin source, when you don't want it's data anymore.
• [BroadcastHub]: Although allows for 0-N consumers, we use it for 0-1, as it plays as a consumer (pulling from the source) while it is disconnected from a graph, and then you can re-attach the BroadcastHub to another graph.

• [MergeHub]: allows for variable amount of producers, which is important when you want to move producers from a graph to another
• [BCH_KillSwitch]: it's a KillSwitch between the BroadcastHub and the MergeHub, and it is how you "disconnect" the PauseableSource from a graph.

The rest of the components are up to the implementation, in our case we have a GraphStage that processes the input from all the sources, and outputs to a Sink which writes to Kafka.