Smarter (automatic) handling of lowTrip and highTrip for nticks=0

I have an axis where I wish to have no ticks. Unfortunately, even no ticks, the scalar factor pushes the margin depending on the use of the \u switch. See the two attached figures.

I'd petition to have a smarter automatic handling of axes with nticks=0. Essentially, I would like to see that nticks=0 would automatically set the lowTrip and highTrip for engineering or scientific notation to ± infinity, thereby removing the (confusing) jump in axis location.

using \u without \u

To follow on this, I have an odd behavior that I cannot understand. The two images below show the results of setting the axis to autoscale nice only in the visible region. One case is with axis tick marks, and the other case is without. In the first case, lowTrip=0.1 and highTrip=1000. In the second case, lowTrip = 10^-20 and highTrip=10^20.

autoscale nice with ticks (330.55 KB) autoscale nice without ticks (306.48 KB)