In other words, you don't care about waveform integrity and you don't need to maintain the waveform's power capability after it's been clipped.
If that's what you want, just put a resistor after the input to the clipping circuit followed by two schotkeys in anti-parallel to the ground rail. They'll clamp the entire signal down to +/-0.5V (you just use a single one with anode on the line and cathode on the rail if it's has zero chance of going negative). Just size the resistor appropriately and it will dissipate most of the heat while limiting the current through the diode and the diode will dissipate practically no power so you don't have to worry about it's size anymore.
That 0.5V margin should give your electronics enough room to detect the zero cross. If it doesn't you can use more diodes in series to increase the clamping voltage a bit or just put zeners back in since the resistor makes it so power handling capability of the zener a minor issue. It's literally the same concept of what the resistor is meant to do in a zener regulator:
http://www.reuk.co.uk/wordpress/electric-circuit/zener-diode-voltage-regulator/
The point I'm trying to make is, use a series resistor. Also, with the resistor the clipped waveform is isolated from the primary waveform so it can continue doing whatever you want (like supplying high currents and power to something bigger).
TVS diodes are much faster than zeners, but like I mentioned in my first post, they aren't built for super-accurate and stable breakdown unlike zeners.
TVS power ratings are based on a waveform, not continuous operation, due to their intended application. That waveform usually is described by 10/1000 or something similarwhich represents timing points in the transient or ESD spike waveform. That's part of why their power ratings appear to be so much higher.