Well, this is not the first time I have seen this topic here. I'm a poor student so constructing adequate models and investing in the flow visualization and measurement equipment is out, but I do have access to a computational fluid dynamics (CFD) package called FLUENT. This is an extremely powerful tool for fluidic design as it allows for not only flow visualization but also measurement of forces/coefficients acting in any direction you would like.

The setup:
I tested a "square" stab with a length of 5" and a thickness of 0.25". This seems to be fairly standard for most model stabs. The model velocity was set to 60 mph at standard temperature and pressure. I varied the angle of attack from 0 to 11 degrees and mapped the associated drag coefficients. The attached images are at angles of attack of 0, 5, 7, 9, and 11 degrees from left to right. The same setup was then used for the same stab with full rounding of the LE and TE.

Results:
The flow separation on the square stab starts at or before 5 degrees and becomes much more severe as the angle of attack is increased. (visible by the lighter blue near the leading edge). Also, as expected the square stab has vortices that are created and shed off the TE (I would suppose that future work would be placed on analyzing the frequency of the shedding and seeing if it is in the audible range which would be pretty cool!). The round leading edge has a huge impact on the separation progression as well as the shed TE vortices. They are actually not visible in these shots because they are extraordinarily small and occur primarily in the boundary layer. I have also included a graph of the drag coefficient vs. the angle of attack. The trendlines are not totally accurate as more data is needed as well as real model validation. As is evident from the graphs the rounded leading edge produces a drag coefficient that is 50% smaller than the square edge.