There's much wisdom (and a bit of controversy) in this thread.
Suffice to say that I fall in the camp that there are many reasons to advocate long arms, but significantly improved tracing distortion is not one of them.
I recently fired up John Ellison's seminal work - his Baerwaald / Löfgren spreadsheet to model various geometries, setups, the relative significance of getting zenith right vs. overhang, and several other scenarios.
If anyone is interested, here's the link: Modeling Various Setup Parameters
Thom @ Galibier