In Reply To:
it just dawned on me that I have wind tunnel data addressing this very question
Okay, here you go: changes in drag* from 0 deg of yaw going out to 2.5, 5, and 10 deg, then back again for the same rider on the same bike in the same position but using a flat disk and a "bump" disk (but also different front wheels):
"bump" disk:
23
-31
-175
-50
-19
flat disk:
-19
-71
-51
-15
40
So, the "bump" disk shows a nice reduction in drag as a function of yaw angle, which is maintained even on the return trip, i.e., from 10 to 0 deg of yaw. When using the flat disk, however, there is less of a reduction in drag as a function of yaw angle, and this benefit is clearly greater when going from 0 to 10 deg of yaw compared to the opposite. (IOW, with the flat disk wheelset it appears that once flow is detached it is unable to re-attach.) These results are consistent with Zipp's claims that adding a bulge just below the brake track makes a wheel less likely to "stall".
*Data in grams, body axis model.