I just did a very quick & dirty test of the same high contrast subject with vertical lines - a window in my kitchen
. I panned quickly left and right.
The same framing of subject using the ETC mode actually showed less jello-cam effect than the normal 'full-frame' method. Both instances showed some jello-cam but the full frame was actually far worse. I was of course intentionally trying to create the effect.
Thinking about possible reasons for this, it strikes me that there may well be less scan lag over the 1920 x 1080 recorded reduced area of the sensor in ETC mode and far more over the full size of the sensor.
FWIW - Certainly my test showed the opposite of your observation.