Hello,
when you read the Cyclone III handbook's configuration chapter, you could wonder, if the device is 3.3V tolerant at all, but from the maximum ratings, this is the case without doubt, including JTAG and configuration interface. Comparing Cyclone II and III maximum ratings, the margin for voltage overshoot versus time is clearly smaller, but the absolute maximum ratings are nearly unchanged. Thus a configuration circuit that could damage Cyclone III devices would also be harmful to Cyclone II and others, usually operating with 3.3V powered configuration, provided the datasheet values are comparable.
But a protection circuit for an external connected configuration interface is menaningful anyway, e. g. a 100 ohms series resistor with a dual schottky diode connected to GND/VCCIO. It should be able to guarantee the allowed AC and DC ratings for I/O oins, except with ESD pulses.
To my opinion, the new thing with Cyclone III configuration interface is the VCCA dependency and the issues arising when VCCIO is powered after VCCA or rising non-monotonically. I guess, Altera developers have been somewhat suprized, how much confusion at customers side could be caused by some simple design changes. Afterwards, they took the easy way and ordered that Cyclone III configuation should be powered by 2.5V VCCIO connected to VCCA, exclusively. This, by the way, would also reduce the danger of overshoot induced damages.
I think, this could actually be a good advice for completely new designs. But customer may have different motives to continue with mainly 3.3V VCCIO, e.g. continuation of existing designs or device stock. Altera application engineers should rather help the customer to realize their design requirements as far as possible than suggesting arbitrary changes.
Regards,
Frank