darkwave - The problem with that experiment is runner load on the belt, how the treadmill's controller handles different speeds and the machine heating up.
For example, on my machine, if I wear the Stryd while walking it will be absolutely dead nuts on. I mean almost exactly the same distance read over time with no variation. However when I start running, the machine and the motor windings start heating up and the distance and speed start to drift slightly from the target. I really doubt that many treadmill manufacturers invested a lot of extra cost just to dial in the accuracy of the display to the belt speed over the range of speeds with any person loading it at their various cadences and ground contact times. It's just not that "tight".
On my particular machine (NordicTrack Commercial 1750), at 9 minute pace it will drift slowly by up to about 10 to 15 seconds per mile off from the Stryd measurement. It takes about 65 to 70 minutes to get all the way to max error where it stabilizes and stops drifting. My heart rate follows it and also stops drifting once the pace reading on the Stryd also stops. It's not just the pace but it's also the Power reading. It all matches.
It's slightly worse at 8 minute pace (about 20 seconds per mile off) and again at 7 minute pace (about 30 seconds per mile off). I assume this is from greater heating. The same is true at high speeds for intervals (over 30 seconds per mile off). One could probably plot the %error vs speeds over time. Again, these are loaded conditions with me running on it with my weight and ground contact time. And at low speeds where the machine remains cool, there is no drift at all. And the speeds are spot on right at the start of the workout.
My point is that there are a TON of variables affecting the actual belt speed over time. In my experience the Stryd pace and power feedback track with my heart rate. I believe a great test to check that accuracy is to either wear a heart rate monitor and compare to outdoor conditions OR repeat your test but do it with a runner on it over a variety of steady speeds over time curves. At appears to me that the Stryd is consistent and accurate. It could be that my pod is good and yours is bad though (as you mentioned)