To go back to an earlier post from Jim (MacColl reassessing his approach to singing a long ballad after research on listeners' attention span) - I knew of Jeannie Robertson's ' Harlaw' long before I ever decided to "sing it out", and when I did I found the 20-22 verses and choruses overlong. But I didn't want to lose any of the narrative, so my first choice was to sing pairs of verses with one chorus - which worked fairly well because the first section of the song is question and answer between various combinations of people. Having made that choice, I found that as I worked it up to performance level the tune was changing in subtle ways:- for example, during the questionings at the start the tune might stay within a restricted range (missing out the high notes at the end of line1 and at the start of line3) but still in what I thought of as a major key. When battle was joined, these high notes quite readily reappeared and were useful in adding drama and energy. Then I realised that as the narrative progressed and Macdonald was slain, some verses were naturally falling into a more modal/minor tuning. (Apologies for being so technically ignorant!) But I would like to say that these choices were not then set in stone, and have varied in other performances - probably dependent on audience response as I was singing.
And audience response is a great thing - there was an occasion when the late Norman Buchan brought Jeannie Robertson down to Glasgow for a concert and he took her out to our school (Rutherglen Academy) the next day, where she sang 'Harlaw' to his second year English class. When he later asked the pupils what they thought, a 13year-old boy said,"See, Sir, battles must have been awful noisy in those days!" ['An' ilka sword gi'ed clash for clash']