Thanks, everyone, for the replies!
+1 for <noteGrp>--that seems like a good solution to me. To speak to Syd's comment: I do think these notes need to be grouped, since they refer to different levels of granularity of a particular note. There is, really, only one annotation, but with different levels of explanation that are aimed at different audiences. (I also think that, in terms of economy for encoding, having a wrapper is far easier.) And, if <span>s can be grouped in <spanGrp>, then surely <note>s should be too?
My only question regarding <noteGrp> is about attributes. Would it be @target and @targetEnd like with <note> or would it be a different mechanism? The issue with that in my mind is this bit of prose in the GLs:
>This attribute [@targetEnd] is retained for backwards compatibility; it may be removed at a subsequent release of the Guidelines. The recommended way of pointing to a span of elements is by means of the range function of XPointer, as further described in 126.96.36.199 range().
If that's the case, then it seems odd to make a new element with an attribute that's not actually recommended. But I don't think it makes sense to create a new @targetEnd that is different from <note>'s @targetEnd. But it also seems like extra work to create an attribute class for an attribute that's not recommended anyways! And it also seems odd to recommend XPointers when, as Hugh Cayless points out in his JTEI article, they suffer from a lack of implementation . (FWIW, my vote would be to harmonize span/(@from|@to) and app/(@from|@to) and then add <note> and <noteGrp> to that attribute class, but that's perhaps a slight digression from the question at hand.)
: Hugh A. Cayless, "Rebooting TEI Pointers", Journal of the Text Encoding Initiative[Online], Issue 6 | December 2013, Online since 04 September 2017, connection on 31 October 2018. URL : http://journals.openedition.org/jtei/907 ; DOI : 10.4000/jtei.907