After that, the tests for the randomized part still go wrong on Plone 4,
where the same code works in plonenext 3.3. I guess something has
changed in the LazyCat internals, and the code in the part of the
portlet that does the randomizing depends on those internals, apparently
for speed reasons. It even has a comment: "Kids, do not try this at
Anyway, as far as I see in this test and when testing it manually in the
browser, the result is that on Plone 4 you never get a result when you
use a randomized collection portlet.
Fixed in http://dev.plone.org/plone/changeset/45917
If someone knows a better way that works in both Plone 3 and 4, and
makes it (apparently) faster again on Plone 4, that would be nice.