Galapagos - Runs ok with Disabled components, but...

I often insert "switches" into my definitions, to "break" them at certain points to allow for snappier editing of parameters that are in the the beginning without affecting the entire chain.

(Ie. edit a spline that will get copied, extruded, subdivided, etc is very slow, so I break the flow before it gets extruded using a stream filter to edit the spline without all the subsequent actions).

This typically works very well, but for some reason with Galapagos, that's not the case.

Even if 90% of the definition is empty (because a base geometry is not being passed), Galapagos is very slow at testing the solutions.

The work around I have found is, I select the part of the definition that is already "empty", and disable all components. After this, Galapagos solver runs very fast.

So my question is: Why would a bunch of empty components (but not disable) slow the solver down more than fully disable components, even if the genome parameter do not affect the them?

See attached screenshots.