There is a Certification issue here. You must be able to delay action for a certain period of time before the situation becomes critical. This 'intervention' time is made up of the pilot's cognition factor and the time it takes him to react. Typically this is assessed, for certification purposes, at 2.5 seconds for the engine failure case. So less than this is cutting it fine. I have heard tell of delays longer than 1.5 seconds in the R22, though personally I wouldn't want to wait even that long. A word to the wise, PLEASE, don't even think of trying this for yourselves. Professional test pilots and flight test engineers have taken the risk on your behalf!
The best protection and cushion against catastrophic loss of RRPM is anticipation and having your hand on the collective whenever possible, but certainly if you are climbing. (Worst case: high blade pitch angle and longest delay in establishing rate of descent airflow.)
There are helicopters that have much more benign rotor decay; the B47,B206 and other machines with high rotor mass and thus high inertia.