I think the answer is that all time "counts" as far as energy expenditure is concerned thus the higher the sampling rate, however "noisy", the better. Given the shape of most yaw distributions we have and the fact that most samples have been taken at ~1hz (iirc) I think we can safely conclude that the resulting yaw distribution skews low. Maybe not much, but at least a bit.

I'll throw this out there: Josh Portner (Silca/Zipp) recently let an interesting tidbit slip. Zipp did some testing where they went from 30 degrees to 0 degrees to see when the flow would reattach and they found that reattachment did not occur at the same angle as detachment for the rims they tested. Instead they found that flow didn't reattach until a lower yaw for most rims. He also hinted that shapes that performed better at higher yaws saw their flow reattach sooner when going from 30 to 0 degrees. For most modern rim/tire combinations we see today flow seems to detach around 12-15 degrees. For all the modeling that's been done (that I'm aware of) it has been assumed that a given wheel's performance from 0-20 degrees is the same as its performance from 20-0 degrees. Separation is admittedly rare for most riders but, again, this is a consideration that might be biasing us toward low yaws in design and equipment selection.

Something that has always bothered me, personally, is that we test in the tunnel with the front wheel perfectly aligned with the frame when, in reality, the wheel is constantly moving a few degrees back and forth from zero. If I had the time, I'd figure out a way to hook up a servo encoder to my headset to gather data.

...rambling here....

If we think about what happens when a side wind hits a rider, the wind tilts the rider/bike away from the direction of the wind which causes the front wheel to turn back into the wind. The rider then counteracts this and resets the system. In this instance, if you had two different yaw sensors attached to the bike at two different places (the handlebars and, say, the top tube) you'd get two different readings for two different parts of the system. If I'm thinking about this correctly the frame would see a higher yaw than the front wheel. The unknown for most of us is how the front wheel down tube system works when the two aren't aligned. It would seem to me the most pragmatic approach would be to include the steering deviation (1, 2, maybe 3 degrees) from center and add that to the yaw distribution that the wheel "sees" as the goal is for the flow to stay attached to the front wheel and then transition smoothly to the down tube.

Wrapping up my ramble here...

In light of the forgoing, perhaps we should be looking more closely at the performance of our equipment at higher yaws.

Edit: this might also make one reconsider tire selection as some fast rolling tires really hinder the performance of a wheel at higher yaws. The Turbo Cotton and GP TT come to mind.