Dear fellows, 

I am bringing to your attention an issue from a student. She is preparing the edition of a medieval chanson de geste. The encoding of this work in vers is very straightforward, it uses divs, containing <lg>s, containing <l>. 

The several witnesses have a common trunk where manuscripts only show minor variants. Then, for the ending, the tradition divides into two branches, each with an alternative ending.  
The issue is, once again, that for no reasonable reason <app> is phrase-level, while it really should be at least paragraph-level. As a consequence, the valid encoding would be to create an <app> for the content of each <l> (not even <lg>), which is nonsensical on a philological point of view. 

While we wait for the much-needed upgrade of <app> from phrase-level to paragraph-level (or better div-level), what would be the best-practice recommended to encode this alternative ending? 

Best regards,