The way I have resumed to myself the 2D plot is about like this:
A real data base system, or virtual depending upon the plotting arguments x, f(x). It has auctionnering search so that if one trace wants to plot but in the data base one does not [like Gamma(3,x) that starts at 0] it will force the defective plot to the first value, for instance it is possible to extend Gamma(3,x) into the negative region.
The tracking tool goes further than root, it goes lower than Given/Find. The autoscale is a wise math extra as it gives bounds (most of the times). To the contrary of many complains, the Mathcad QuickPlot 11 and lower, is a lot more useful and automated package than Mathematica for example. Detail in your sheet: replace NaN by the blanking tool " ... just writing as reading my recollection. The limit in points plotted is the limit of the data base 499999 points, i,e: one (1) trace 499999 points or 499999 traces (family) an one point each ... not too much use for that. There are inaccuracy in the pixel division of the plot [Remember Richard demo]. Quickplot is Metafile, you can thus copy and paste as metafile and superpose a bitmap, even a live graph.
What else do I know or don't recall ?
Sure, I have some feature in head !
jmG