Hello there.Happy to answer.
Tubes in general have lower distortion , the higher the load is.Correct.
In OTL designs, the tube "sees" a load (the speaker load),divided by number of tubes. If the load is halved,then the number of tubes must be doubled, if we want for each of the tube in the output stage to see the same impedance as before.
However
A transformer is an impedance matcher.The load that each of the tube "sees" is the impedance of the load multiplied by step down ratio squared.
For example, a tube in a SET output stage, will "see" a 8 ohm load as 5000ohm if a 25:1 step down is used. If the load becomes 4ohm, and if we keep the same step down ratio of 25:1, the reflected load will be 2500ohm.
In that case,obviously the tube will suffer from reduced power and higher distortion level.
Now,if we increase the step down ratio by a factor of 1.4 ,the reflected load will be again 5000ohm.Tube is happy again,delivering same power and same distortion as in the 8ohm load.
Now,how this is done( the increase in step down ratio of the OTP) , is where the whole game is.Sub par solutions end up with lower bandwidth and/or higher copper losses.But there are some techniques/solutions then bend around these issues.
A tube stage loaded with a step down transformer has absolutely no problem delivering extremely low distortions and high power even down to 0.1ohm, if the transformer is designed for that purpose.
Cheers
Stavros