... Then worried a bit as it was only doing 10GH/s avg for some time.
Reboot the S3 via SSH (not sure whether this works via the GUI).
If you noticed when this happened, it also had an uptime that started before you restarted cgminer. This happened to me (and I was able to replicate) after installing cgminer 4.6, but a reboot via SSH "clears" the relic numbers. the downside to not having a clean slate for your circumstances is that you will not have accurate uptime hashing speed(s).
I agree but find that stopping/starting /etc/init.d/cgminer gets a clean time and doesn't cause the ssh to disconnect.