I had an older Tacx Bushido (the older one with no flywheel) and often (even in the TTS4 software) had the same problem. In TTS4 randomly I would have a ride where the downhill was essentially freewheeling (about 1 in 10 rides) while most of the time the rides were making me pedal hard on downhills to not coast to a stop. In Zwift it was less like this, but still the downhills do not feel like real downhills.
I have since gotten a Tacx Bushido with the actual flywheel, and it reduces this problem but does not eliminate it. The original Tacx Bushido with no flywheel was effectively impossible to calibrate as it had no spindown at all and the TTS4 software failed to be able to run the calibration at all 95% of the time. The flywheel version lets me do a TTS4 calibration, but as others have reported in the TTS4 forums the TTS4 calibration of even the newer flywheel Tacx Bushido does not work properly and (I can’t verify) does not match at all with an iOS based calibration. Depending up my “calibration” of the Tacx Bushido I can find that downhill take a huge effort just to keep moving, or is more realistic. The lighter the clamp, the more realistic.
So… At least on the Tacx Bushido it clearly relates the so called calibration that one does in the Tacs software (for the the TTS4 on the PC). I understand that Neo does not have to be calibrated, but nonetheless it is still some component of Tacx software doing the same thing in both cases. I would blame it therefore, given my Bushido experience, on the Tacx software and/or Tacx driver that comes between the Neo and Zwift.
My suggestion would be to experiment with a) changing rider weight (up a lot and down a lot) to see if that impacts it b) changing % slope in Zwift.
I had some settings at times where depending upon a mix of the supposed Tacx Bushido calibration and rider weight and slop that had my wattage uphill reported by the Bushido as maybe 2/3 to 1/2 of my actual power and then cresting a hill and going downhill I would pedal with little pressure and my power would be reported as nearly double what I put out on the uphill. That all seemed to relate to the Tacx software/hardware mix reporting the wrong values to Zwift more than anything. However, Zwift is also feeding things back to the trainer for resistance and seems not entirely blameless.
At least for my newer Bushido, I have finally been able to some reasonable compromise out of it by playing for a long time with the notoriously wrong (at least for Bushido) calibration, rider weight and slope. It has been a long fight to something close to acceptable. Perhaps with the Neo there is something internal that is not a parameter you can see that is locking you into one of the bad scenarios I had for years on the Bushido and in particular the old Bushido?
Is this common to all Neos?
Can other riders with Neos input the same parameters (rider wt/ht and slope %) and then report if they see the same problems?