However, the list of x elements that it finds is saved as a kind
of index, linked to the document node, so that the next time you do //x on
the same document it already knows the answer: which is therefore cheaper
than repeating the /a/b/c/x query. This also applies if the //x is part of a
longer path expression.

I'm not sure I understand what you said here? Does this mean that
contrary to common wisdom and understanding multiple uses of //x will
be more efficient than multiple uses of /a/b/c/x  when using Saxon?

Cheers,
Bryan Rasmussen