I have some experience with LCS and 7970.
So far, everytime the GPU overheated it would crash and a reboot would be required. In that state the GPU didn't produce heat anymore.
You can do the test yourself and pull the plug for the pump while the GPU is mining.
Incrementally wait 30sec and put the plug back. This way you learn how your system reacts in case of a pump failure
I expect that the GPU will crash if the water gets too hot and that's it. Then you can wait and see how long it takes for the heat to dissipate.
BTW it helps if you overclock a bit. It makes the GPU more sensitive to high temperatures so it will crash earlier
I have two pumps, but only because I need them
(default pump wasn't strong enough)