Hi Sigfrid, the problem with square roots in groff's eqn is that groff only knows a single type of square root (the small one). Therefore if an equation requires a higher square root, the only thing groff can do is set the small one in a larger size, and you get an overly thick, overly wide symbol as a result. This is demonstrated in your squareroot-problem.pdf.
The proper solution to this, IMO, is to use fonts which include multiple types of square root symbols and to make eqn use these symbols when they are available. This is what TeX does and it's also how neateqn has approached the problem. Cheers, onf
