Nothing inherently, but the problem is that it can be misleading.

Let’s make an arbitrary example. Say a study is designed to measure the difference in lifespan between group A (water drinkers) and group B (coffee drinkers). Imagine there is a true difference in lifespan (say, 87.8yr vs 87.9yr) and the study is powered to observe a statistical difference.

Now imagine plotting the average lifespans between the two groups.

If the scale starts at 0 and ends at 100, the difference would be imperceptible.

But if the scale is shown on a range of 87.6-88.0yr, with breaks of 0.05yr, the difference between the groups would appear quite large. This Y axis scale is shifting your frame of reference to potentially think the difference is bigger than it actually is.

Maybe that 0.1yr is meaningful for some contexts or opinions, but maybe it’s not. This is why reading the units on an axis is very important: so we’re not mislead by the visual difference between two group values.