JAE,
I follow the point you make and I agree that the FAILURE MODE of a dowel is based on concrete shear however, what about the case of load transfer dowels in a floor slab? They transfer the load via dowel shear and don't have much developement length, particularly when they are either a smooth dowel or a greased dowel.
In the case of a footing, many continous footings are bullfloated off, thus little surface roughness. Or how about the case of a shear key in the footing, which works in shear only?
I guess my point is, as I'm sure you know, the ACI code still has many confusing and CONFLICTING provisions (the newly published ACI Masonry Code is another good example). The connection of a "normal" foundation wall to a continous footing is VERY seldom in tension, it is in compression from the dead weight of the wall and loading from the structure above. That is why, in my opinion, a straight dowel in a continous footing is OK. Now, dowels from a heavily loaded concrete column into a spot footing are entirely different. They could be pushed through the footing so, a straight bar pushed into the footing is not acceptable. The footing has to be investigated for punching shear.
This discussion started out asking whether "stabbing" of rebar is acceptable practice. I think, in MOST cases yes, for the reasons I've stated above. I do agree, that in some cases, it is not acceptable. Design knowledge and field experiance tell us the difference.