We can observe various non-linearities on real-tubes graphs. Amplification factor ยต changes the least on various "points", and anode resistance rp more. But, thought that SE amp varies its Zout and DF wildly during the work isn`t correct. If you measure DF on say, 1W, 3W and 5W and DF on say, 100Hz, 1kHz and 10 kHz you`ll get about the same result - about 3 in my example. If your theory is correct, then you`ll get very different results.