Thank you PZ.
What I mentioned earlier today has nothing to do with CFD and lets leave other dirrct aerodynamic disadvantages aside. As an engineer, I'm sure you'll agree that a longer cantilever is far more prone to deflection than a short, fully braced one. I'm also sure you'll agree a floor is actually far from a cantilever, as it is not attached to the tub along the entire circumference, allowing for a lot of play where there can't be any with an actual cantilever.
For any given load case, you "fight" the stress with geometry and deflection depends on the material and stress. So to reduce stress you can change material (so use steel, which is too heavy) or increase the rigidity with geometry. In this case - make the floor thicker, a lot thicker. How does this affect aerodynamics? Do the rules allow for such thickness? How big is the weight penalty?
Substantial floor deflection can and does make things very difficult for a driver, having an unpredictable car due to unpredictable aerodynamic behaviour. This is very evident with Mercedes this year, on every race, hard to set up and hard to handle. If they would choose to change shape of the floor again, for more predictability, that could only be achieved by giving up some downforce. Their cureent lack of cornering speed can be solved, at least to some extent, once they have a more predictable car. The rest would have to come from bigger rear wing.