So I decided to spend today’s Round Bald workout collecting some data 
Conditions for this test were:
- Rail trail with no discernible grade except one bridge that rises over a road
- About 11 mins out, 10 mins back on same course; broke into two laps to separate turnaround and most of the time required getting back up to speed
- Each out and back was 240 +/- 1W average, and NP/AVG were always within 1W - very well paced to my plan
- ~38F - full winter kit - face partially covered, thick winter gloves, Gore shoe covers, jacket - well fitted, but miles from a skinsuit!
- Nothing stripped off bike - two bottles, frame pump, soft bento tool bag - even had very non-aero bell
- Was lucky that it was quiet - never once had to break position or use brakes so data should be pretty good quality
- Used lap timer between visual setpoints to get time/distance
- Probably lost some weight in each test from sweat/vapor loss
- GPS data - my speed sensor magnet is broken
- Road bike with the redshift aerobars, forward seatpost
So with that - I was sadly disappointed by my new P09 helmet 
Or perhaps the Ballista is really that good? I also expected more of a penalty for the IRC file tread x-guard tires over the GP5000TLs on carbon rims.
I used bikecalc to basically dial in the parameters to line up with my baseline run, and then modified the wattage to see from that point, how much it estimated I would have needed for each speed increase on the 2nd and 3rd runs.
I’m also going to try re-analyzing on strava segments more inside the run as a second check against any differences in speed transients for the first minute at each end.
The winter clothes definitely cost me close to 2 MPH so could be interesting to repeat at higher speed - but overall, the differences are probably less than what many people would expect.
Anyway, thought it would be interesting to share and maybe others have some ideas… haven’t gotten around to trying to back a CdA out of it yet.