I agree that if metadata requirements are met, the <group> proposal seems
most natural. Yet, this seems to imply that the transcription has to occur
at collection level. Since only full <TEI> texts can be validated, there's
no way to validate single poems (ie. <text>s)? 

>At the level of encoding management,
>there is no reason why a transcriber-encoder shouldn't work on a document
>instance where the <group> has only one <text> member if that seems

I'm afraid your hint isn't clear to me...

>Perhaps things would become clearer if you could say more about what you see
>as the issues behind
>> the encoding of the same texts at different levels (as autonomous texts
>> AND parts of a bundle)
>and in what respects <group>ing of <text>s is inadequate to meet what you
>need to express.

Well, some of the poems are unique; others occur in different versions. We
intend to create one parallell-segmented collated version for each of the
latter poems, thus 'unifying' different <text>s. A collection would then
consist of the unique poems + the unified versions. In this respect, it
seemed easier to think of the constituting parts as <TEI> documents. I must
confess, however, this is still at the conception phase, so I might be on
wrong tracks.