PaulK: "Why don't you do it?"
Its quite enough that you wish not to run your model in the Ensemble for validation, declaring that path "bad advice". Good advice is not to futz with poorly documented GitHub code. In any case, your calculated predictions for the 2280 window will eventually serve the purpose.
My interest is to continue developing the alternative Signal Crosstalk hypothesis to strong Lunisolar Forcing by literature research and heuristics. If its possible to well-enough specify and quantify such noise, running your code may be redundant. If you can falsify the alternative hypothesis, that's good too.
To recap: Your ENSO-QBO Lunisolar Forcing hypothesis is at odds with Chaos Science peer orthodoxy. I concede upfront there must some forcing hidden in the statistical chaos, for as Dirac put it, "pick a flower, move the farthest star", but its unclear how much of the effect in the data is Lunisolar Crosstalk and how much is your Forcing, given the nature of the multi-sensing.
This alternative view strikes a nice balance between the two geophysical camps. What is left is to determine the ratio of crosstalk to forcing, and filter the crosstalk from the Ensemble Model to improve it.