> Given this XML:
>
> <xsl:variable name="foo" as="element(foo)+">
>   <foo/>
>   <foo/>
>   <foo/>
> </xsl:variable>
>
> I seem to remember David C mentioning something about using
> as="element()" means the elements are no longer siblings...
> but I can't find anything about it now.
>

It's not a question of being "no longer" siblings. You've simply created
three element nodes. You haven't attached them to any parent, therefore they
are not siblings, and never were. If you want to make them siblings, use

<xsl:variable name="foo" as="document-node()">
  <xsl:document>
    <foo/><foo/><foo/>
  </xsl:document>
</xsl:variable>

or use the old syntax, which you can regard as an abbreviation for the
above:

<xsl:variable name="foo">
   <foo/><foo/><foo/>
</xsl:variable>

Yes, thanks, it's becoming clearer although I'm still having issues
with a "document" containing three root nodes that are all siblings.

<xsl:variable name="foo" as="document-node()">
  <xsl:document>
    <foo/><foo/><foo/>
  </xsl:document>
</xsl:variable>

...is this because "/[2]" was thought to be impossible, or a 1.0
compatiblity issue, but /[2] certainly exists here.

Ultimately this all stems from a desire to give every variable and
parameter an "as" attribute to take advantage of the "fail-early"
policy of XSLT 2.0.  In a multiple pass setup I would normally use
element()+ on a variable containing rootless elements, but have ran
into this sibling issue*.  I've switched to using document-node() but
have now hit this slightly non-intuitive part of / having multiple
children.

*It would be a useful feature if Saxon gave a warning if someone uses
a sibling axis on an element in a rootless tree...similar to when the
context node is an attribute (if that's feasible)

cheers
andrew