The logic to this feature should probably be reimplemented, or even better, redesigned to be more useful in the general case. Someone might want more than a "left" and "right" Y-axis -- matplotlib will give you as many as you ask for with twinx(). It might make more sense as a per-spectrogram or per-trace attribute rather than per-panel and only via pseudovariables. For example, consider a tplot variable with three traces r, theta, phi, where theta goes from 0 to 360 or -180 to 180, phi goes from -90 to +90, and r is in some distance units. You'd want 3 different scales for that, but there's no way to make that happen with tplot as it stands now.
The logic to this feature should probably be reimplemented, or even better, redesigned to be more useful in the general case. Someone might want more than a "left" and "right" Y-axis -- matplotlib will give you as many as you ask for with twinx(). It might make more sense as a per-spectrogram or per-trace attribute rather than per-panel and only via pseudovariables. For example, consider a tplot variable with three traces r, theta, phi, where theta goes from 0 to 360 or -180 to 180, phi goes from -90 to +90, and r is in some distance units. You'd want 3 different scales for that, but there's no way to make that happen with tplot as it stands now.