Plots take multiple limit finding iterations to resolve

[asinghvi17]

This is because the dataspace coordinates change on limit change, which causes a change in limits.

We could define tightlimits or so but I don't think it would change the fundamental issue - something has to cycle back to the axis/scene.

[jkrumbiegel]

The precise x limits of a swarm plot are screen-space dependent, so I don't think that info should go into the actual data limits being reported. I would either report only the limits of the x coordinates passed in, or maybe expand that by half a category because you'd expect that people set the plot up so that adjacent categories don't overlap.