I've generally been running "joyride" development builds on my XO, and reflashing back to the original configuration every couple days to start experimenting on a fresh base. The re-flashing (this is like Ghosting or cloning a machine for you school IT folks) from USB works well once you get it going, although it seems to take 1-10 attempts to for the laptop to properly detect the USB key, so if you're trying that, be persistent.
Anyhow, authenticating to my WEP-encrypted network (yeah, I could turn that off, but I'd have to figure out what the password to log into the AP is first) works well for me in the shipping configuration, less well in newer builds, which seems consistent with this bug report.
It is worth noting that this and other outstanding wireless problems may have their root in this meta-bug: wireless firmware not in house. That is, the software that runs the wireless chipset is closed source, for this reason. The upshot is that for the mesh networking hardware to keep moving packets when the rest of the laptop is shut down, it has to run its own little operating system, the one Marvell uses for this purpose is proprietary, and OLPC hasn't finished an open source version of the whole stack yet. So it is one key place where neither OLPC, you or I can fix bugs. They can only ask Marvell for help.
Also, handling secure WiFi connections is not a requirement for kids in the developing world using the laptops, at least not an immediate one, so it is unclear how one should evaluate the severity of this bug in general, although it certainly is annoying me.