It’s not just context.

It is similar to paralysis - in that now every prompt the model has to reason over more tools to possibly decide to use - this is surely a deviation from training the more tools you add