Hello,
i have experience sailing cats in strong wind (no spi though).
for me the idea of over sheeting the sails downwind is theoretical and anti productive. as sail600 mentioned keeping up your speed is critical, especially in the jibes. jibbing is when a capsize is most likely to occur in a blow.
steering with the waves and surfing is crucial. usually you don't go dead down wind(DDW). you reach up a hair and sheet in a bit. then you can bear off in the big gusts and even sheet out. technique being the difference on weather or not you go over. with no spi. it's ok to head up when you get to the wave in front of you. if you don't you will auger into it and possibly pitchpole.
for me the traveler and barber hauler are all the way out. sails sheeted enough to avoid luffing except in the big gusts. do your jibes on the top of a wave. if you match the wind speed with hull speed the boom won't slam that hard. after it goes acrooss, the boat will shoot down the face of the wave. deliberate and percise tiller action required. also pull the boom down on the jibe to reduce the slam effect.
without the spi. you wouldn't be out on the wires. then you don't have to reach up so high and can go DDW in the biggest gusts. also it easier to pitchpole whilst flying a hull because all the pressure is on the leeward hull causing it to be further underwater. if you get a big gust with the leeward hull already depressed there will be less bow above water when you go to bearoff making pitchpole a greater possability. if the bows dig in and everyone holds on and sheets out you could save it.
these ideas are for a cat w/no spi. sailing in the ocean in winds above 20 knots on various cats ranging from H16 to tornado.
you may not get to the mark first if you don't fly the hull downwind in high wind, but surfing the swells with both hulls mostly in the water is quite fun. whereas jumping over the back of a wave whilst on the wire downwind and augering in at the bottom doesn't sound that appealing to me.
please pardon spelling and punctuation errors. -dr