The position and thickness of the overline rule are not consistent with the top-right hook in radical (top) glyphs that have such a hook. The position differences may be due to rounding in FillRect or aligning the glyphs to device pixels. The thickness differences are due to the overline rule thickness having little to do with the radical (top) glyph. For fonts that contain appropriate glyphs for the radical overline, a stretchy nsMathMLChar may help things. For other fonts, using the vertical extension glyph instead of the font's top glyph with hook could be used for the radicals stretched by parts. The overline position would need to be based on the right bearing of nsMathMLChar (- 1 pixel) rather than the advance width.

When present Microsoft's MATH table in OpenType fonts (bug 407059 comment 0) can contain a RadicalRuleThickness value.

Created a webpage that shows the bug. First 4 roots have bug last 4 roots do not) https://eyeasme.com/Joe/MathML/mroot_bug.html image provided is of the created webpage.

Firefox 4.o seems to have the right hook problem solved, and mostly looks great! When the web page given in Comment #3 is zoomed in and out the overlines sometimes vary their thicknesses as the image provided shows.

Here the website given in Comment #3 is zoomed in showing that sometimes the overlines are of different thicknesses.

What I called Comment #3 is actually called comment #2, because the first comment is called a description and not a comment.