A much better solution is to simply limit the RMS output of the amplifier. How? Well, the absolute simplest way of doing this is restrict how high you can turn the power amp up by putting an appropriate value resistor in series with the volume pot (two for a stereo amp)...
...
barefoot
Barefoot Sound
Sorry for dredging this old thread up, but I could not let this error stand uncorrected.
The 'rms' limiting you describe is actually peak limiting that will still result in a clipped signal (at the output of the preamp) under feedback conditions.
What you have drawn is similar to a power soak in the preamp stage.
The whole notion of harmonic content destroying tweeters depends on a clipped waveform and a passive crossover to route the harmonic content to the tweeter. Your solution addresses neither of these factors. All it does is reduce the magnitude of the clipped signal similar to that zener.
The one thing that you are correct about is that the dynamic headroom will decrease. The same could be accomplished by using a smaller amplifier.
Speakers must be able to reproduce peaks cleanly and that means no clipping on their signal is allowed at any point in the signal path. Even trying to limit just the tweeter voltage with a zener will put an unacceptable strain on the crossover while trashing the sound.
The proper way to prevent feedback from blowing drivers, or any other signal from blowing drivers either, without causing objectionable distortion in the processing, is to use compression. Compression distorts the signal also, but in a more benign and tuneable way that generates fewer harmonics that sound bad and blow tweeters.
Modern powered speakers tend to use soft limiting protection that is basically a form of pre-tuned compression, and/or to use discrete amplifiers for each frequency band so that the power in each band is inherently limited by the amplifier capability.
Such limiting can also be used with passive crossovers effectively. Typically the configured compression has an attack time that lets dynamic transients through while preventing sustained feedback from frying the tweeters. Compression is the only practical way to limit the average power without also limiting the peak power and destroying the dynamic range.