I found this interesting:
open the good and bad log files with Excel (take the defaults on the import wizard) and delete the first 3 columns (otherwise, the compare will show up every line as different, since the first 3 columns contain information that is different from run to run, such as the time and the process id). Finally, compare the resulting log files (for example with WinDiff

He massaged his text logs into usable form using a spreadsheet! Instead of running a text utility to pipe input and output, parsing out the columns, and rearranging them, deleting them, sorting the rows, or whatnot, he used the spreadsheet's ability to parse the text and simple features of the grid-like system to manipulate rows and/or columns. I never would have thought of that. Maybe that's inspirational, and I'll remember that next time I need to sort on something other than the leftmost characters.