How long a STABLE gyro can be kept in low/zero G without losing a critical amount of RRPM depends on the design (is this becoming a familiar theme?).
A Bensen rotor is one of the worst in this regard. It is draggy (because of lap-jointed skins, very low mass, open blade tips, gaps in the upper skin segments, lots of reflex and a sub-optimal airfoil). All of these features make this rotor lose RRPM VERY fast when the disk AOA is reduced to near zero. The only feature of the Bensen rotor that compensates a little for these shortcomings is that it is designed to operate at high RRPM (over 400). If you slow it down to increase its efficiency (by increasing pitch or lengthening the hub, as we all used to do), then you have made it really easy to lose enough RRPM in low G to prevent recovery.
At the other extreme, a rotor that holds its RPM for a relatively long time will have: an accurate NACA or other professionally-developed airfoil; as little reflex as is safe; absolutely smooth surfaces; closed blade tips; and a lot of mass (especially at the tips). There is quite a dramatic difference in RRPM retention between, for example, Skywheels and Bensens.
I think that an airframe design that makes it difficult to sustain zero G is the most important safety feature, though. If the craft noses up when G's are reduced, it will automatically increase the rotor disk AOA and therefore G loading. In effect, the frame will tend to override a sudden forward stick input or downdraft. Again, this feature is easily designed into the aircraft using negative H-stab incidence and/or a slightly low thrustline.
The opposite extreme from the frame that noses up in zero G is a frame that noses DOWN. A HTL gyro (including one that has a H-stab but the H-stab has no incidence or is too small) will nose down, at least a little, when the G's are reduced. A PPO occurs when this nosing-down reaction becomes self-sustaining. That, in turn, takes less than a second -- a fraction of the time than it would take for RRPM loss alone to become unrecoverable.
I understand that a Magni will nose down at first when the stick is thrown forward. If so, that makes sense. However, I also understand that the Magni will stop the pitch-over on its own before it progresses far at all. This also makes sense. The Magni's H-stab is large but has no incidence. The stab won't develop a nose-up force until it has acquired a negative AOA. That means the frame must rotate a few degrees nose-down before the H-stab "bites" and arrests any PPO tendency.
It would very useful if rotor manufacturers would test (and disclose!) their products' RPM decay rates at various low and zero disk AOA's at various airspeeds. But none of them does, as far as I know. The customer is left to do the "research" -- using live ammo, so to speak.