I was a typographer

And I was a fireman. What of it? Because that occupation is equally (ir)relevant to this discussion.

Source code is not prose, or poetry or a novel. No one needs to (nor should!) speed-read source code. Even with prose, a 45/50 character wide column is painful to read. (Tell me you find it comfortable to read this.)

Source code is an entirely different kettle of fish: tokens + structure. And formating source code according to typographical (paper) rules -- which have been comprehensively demonstrated to make little sense when applied to a VDU (by Northern Telecom amongst others as far back as the early to mid-80s) -- is as inane as white on black themes because there seen as retro and thus 'cool'.

For the most part, code only gets wide when indented and the part of the lines you read are only 50-60 chars anyway; but scrunching those 50-60 characters across three 20 character lines, because the current indent level takes you half way across to the arbitrary limit makes the code less readable, not more so.

And no, moving the nested levels out to a separate, called-once sub does not always make sense, especially if it obscures the structure of the algorithm -- which such artificial re-structuring invariably does.