Construct 2 only issues draw calls for objects that are visible and on-screen. Invisible layers are skipped completely and issue no draw calls at all regardless of their content.
In extreme cases (which I guess you might be seeing) a huge number of off-screen objects can still cause overhead in draw calls due to all of them having to reject the bounding box test which checks if they are on-screen. This is a tiny super-fast check though so you will need thousands of objects to see any effect here. This is precisely the problem render cells are designed to solve: it means it only checks instances near the viewport, rather than everything on the entire layout. (But note the restrictions on render cells.)
Invisible layers on the other hand skip their contents entirely (since it makes no difference if they are on or off screen), so it is faster to hide large numbers of instances by putting them on an invisible layer than moving them offscreen. This is a micro-optimisation though, so again is only important with thousands of objects, and render cells should in theory avoid the need to use this workaround.