From what I understand - and happy to be corrected - is that the Sequence node is identical to a regular execution flow. It will wait until pin one has been executed before moving to the next pin.
The only time this isn’t true is with what they call latent functions; delays, etc. It allows you to run a delay on one pin, and then continue to execute something after it.
So for those reasons, the only use of split appears to be:
- Allows the use of latent execution without delaying other execution after the node
- Personal preference - how you like to lay out your node graph
You’re correct in saying you can group slow and fast node groups together to execute them in a specific order. But you can also do that without the Sequence node by putting the slow/fast group before/after the other one normally. So there’s no actual gain there.
Under the hood I don’t know what the cost of the Sequence node is, but if it actually matters, you should probably write that one in C++, not the node graph.