Our procedure is this:
reboot slave first.
reboot master after slave has come back
(these are important because SSG are known to "brick" due to bad memory sticks -- easy fix, but if you are not onsite and dont havt the legacy memory sticks the device is dead)
after both are fresh and rebooted, then upgrade slave first.
if you want to make sure the upgrade does not break anything, then reboot the master or execute command to make the slave the master. this way traffic flows to the new firmware and you can test it for a few minutes (master reboot) or hours (making slave temporary master)
if content with results, then upgrade the master and reboot.
this is the safest way to do it. we have been burned by enough SSGs. if you do this, the worst thing is you have a problem with a SSG but your uptime is still 100%