Interesting question. I pulled up some code and run it through (don't shoot me)

  awk '{print length($0)}' * | sort | uniq -c | sort -nk2

... which produced the following output:

 504 1
 112 2
  12 3
[download]

(504 blank lines, 112 lines with a single char (probable a closing brace) and so on). I turned that into a scatter plot and I see that I have a fair spread of line lengths all the way out to about 75. From there the number of lines with longer lengths drops off a cliff, but that there is a clearly visible tail of line lengths all the way out to 120 or so.

Which does tend to correspond to the way I code. I like breaking things up, such as named parameters on separate lines, but as I'm either in an xterm or a Putty session, I just make the window as big as it needs to be, and if a certain line of code just looks nicer at 120 characters then I'll let it be.