Updates: my doctest failure in citation.pyx is a "red herring". In particular, it's not because of the patches here; it's because of a flaw in how the function get_systems works. Often when I've been testing the packages and patches on this ticket, I've created a new version of sage in a directory called "numpy", and get_systems sees that string in the path name and adds "numpy" to the systems used by the particular command. If I rename the directory to something else, then the doctest passes. So let's not worry about this; at some point, someone can fix get_systems so it ignores the initial chunk of the path name (ignore the parent of SAGE_ROOT, for example).
My doctest failure in matrix1.pyx is still there. For some reason, the "Datetime" type codes are not there on some systems. Any ideas why? I glanced at the install log for numpy, but I didn't see anything suspicious, not that I really know what to look for. Here's the log from my OS X 10.6 machine, which is one place I see this problem: http://sage.math.washington.edu/home/palmieri/misc/numpy-1.5.0.log.
Finally, I have a new version of the scipy spkg, which works for me on all of the machines I've tested on: linux, OpenSolaris, Solaris on x86, Mac OS X 10.6. I'm putting the link in the ticket description, and I'm posting the Mercurial patch -- it's very small.
I'm leaving this as "needs work" because of the matrix1.pyx issue, but I think that's the only remaining problem. Please test the new scipy spkg on other systems to make sure, though.