Then I learned that, for example, some characters would have multiple epithets (grey-eyed, neat-ankled, all-seeing, etc) that in Greek would be selected essentially not because they were the best descriptor, but because they were the ones that would complete the line within the meter! So if there were two extra syllables, the reciter would use one of these, but if there were three extra, they could use one of the other two.
There are a bunch of these formulaic sayings, lines, or even entire little episodes, that could be recited to gain a little time to remember the next part, or speed up the action if the audience was flagging, and so on. The entire form was structured around live performance and improvisation. That was really revelatory for me.
Naturally they lose something in being recorded in static form - retaining other forms of natural beauty and (as hnaa points out in another comment) their symmetry, symbolism and meta-construction. Still such a pleasure to read!