Actually, new things can be added under natural selection if they do not
reduce the success so much as to be selected against; they do not have to
provide benefit. Sequences can therefore be built up through some
suboptimal intermediates. Also, pieces of the sequences may be good for
something. Alice Fulton (in a discussion on the Science and Christianity
list on Behe's book) mentioned that the citric acid cycle is not present in
cyanobacteria. However, all but one of the enzymes in it are present in
one of two "half-cycles", and the missing enzyme is quite similar to an
enzyme associated with one of the half-cycles.
It may be possible to add many pieces at once through gene
duplication events. For example, the translation of RNA into protein is a
very complicated system. There are specific tRNA molecules that carry each
kind of amino acid and help connect it into the forming protein. However,
many tRNAs are similar to each other and apparently arose through multiple
gene duplications before any modern lifeforms had differentiated.
Also, it is possible that modern complex systems have evolved
considerably since their assembly, concealing the evolutionary pathway.
I'm not sure how one could test to differentiate between design and lengthy
evolution.
To me, the fact that everything works and is comprehensible is a
better arguement for underlying design than is looking for something not
yet explained in a naturalistic framework. As far as I can tell, God has
performed most of creation by what we would consider "natural" means, i.e.,
under the constraints He made for the day-to-day running of the universe.