Indeed there is a drift on its transmitter.
The procedure is to check other good passes on other stations and see if the drift is stable. Watch out when you checking as some stations hasn’t adjust their SDRs so the signal may have also the drift from the SDR.
After finding the drift the next move is to go to https://db.satnogs.org and suggest the right drift of the transmitter, after that it will be checked and approved or not.
It looks like the drift is dynamic and for now we don’t have any mechanism to calculate dynamically transmitter drifts. I don’t know either if there is any way to calculate it or estimate it with a good accuracy.
The difference between passes seem to be associated with the Doppler shift, being the directo overhead pass the worst.
CW needs a freq shift of about 600 Hz in order to decode by ear, and the current shift is anywhere between 0 and 30 Hz, shifting higher by some 300 Hz might be a compromise as in bad passes would yield some 350 Hz shift and in the good ones would be more like 1 KHz which is still decodeable by hear.