There are a lot of potential bottlenecks in this process. First reading large Excel files can significantly slow down Dynamo. The way it’s parsing excel files is REALLY inefficient. I am biased here since I wrote an excel package called Bumblebee that is a little faster, but even better would be some solution written around OpenXML implementation.
Another possible bottleneck is geometry creation itself. Dynamo spits out previews and whatnot, and that will slow the whole thing down to a grind. The best way to minimize that would be to actually never generate geometry in Dynamo context so that it doesn’t try to render it. If you were to create these objects in Python, then you could pass them downstream without having to preview/render them, and of course you can dispose of them. That would potentially speed things up a little bit.
Finally Revit itself is slow. Depending on what nodes you use to create the rebar instances, you might be able to gain some efficiency by using bulk family instance creators. I am not sure about the rebar nodes themselves, but i will take a wild guess they don’t take advantage of that functionality. I think the only ones that do that are the AC nodes.