I like this topic!
We discussed this last night in after-hours "beer-thirty" round table. In Top Fuel racing, we have seen the advantages (and disadvantages) of aiming zoomies for thrust. There's real world experience to support it (data as well). Nitro combustion gasses expand in a big way, and continue to expand while in the header tubes, and especially when it gets to meet an oxygen rich environment again outside the header. In a Bonneville application, you can ask George Fields about how much rudder he had in the coupe with the 'E Motor' 4 Banger (Half Hemi). After a quick 'pedal at the 3, he wasn't aimed as originally intended.
To further the idea about the Coandă effect, when I built the last set of headers for the Donovan in #44, we did our best to aim for the wing. Can the faster passing gasses increase the effectiveness of an airfoil? Or does the turbulent air disturb the laminar flow over the wing? I suppose we could 'tuft' the wing elements to find out for sure, but the idea sounds pretty good to me.
As for drag, I tend to agree with the "leaf blower" theory. If there is significant aero drag from headers, it's most likely coming from the solid header tubes, not the plume of gas escaping.