To start thinking about the possible diversity of XPaths with philologically interesting content, we have analyzed the 2,390,070 nodes in the croala-cts-1 XML database and extracted all XPaths descending from the TEI/text nodes. It turns out that there are 2563 distinct paths; the deepest two have 13 levels each (of type text/body/div1/div2/div3/div4/p/table/row/cell/foreign/choice/expan), while the shortest seven consist of (unsurprisingly) just two nodes.
The paths were analyzed by the XQuery map-all-paths-from-text.xq. The results are in croala-cts-1-reflevels.xml.
It is my opinion that in a scholarly edition everything should be citable; in a scholarly edition, everything is of some importance (if it is not, it should be left out). This means that even the longest XPath should be described, and made reachable, by CTS URNs.