OK, the turns ratio is why this didn't blow up in an instant. But, yes, the reference design sounds right. What you have is a circuit with a fairly stiff source, and NO resistance to limit Zener current! Incredibly bad design. The only way this could work would be if they were depending on the resistance of the winding to limit current, but with an ~ 25:1 turns ratio, my guess is the 4 turn winding has a fraction of an Ohm resistance. You'd have to figure out how much extra drive it has to know what resistor to provide. The drive on the PFC extra winding is going to vary with line and load changes, such as when the chlorinator is actually dispensing. But, you HAVE to add a resistor or the Zener will keep blowing.
Jon
I would check your 2x 100 k R46 R47 resistors, the c41 c39 c40 capacitors and the 2 In4148 diodes
I would replace R46 and R47 with higher wattage parts to get higher voltage rating in case they are occasionally letting a blast of the 380 V through. They likely used the two in series to increase the resistor voltage rating of the "compound" resistor formed by the two of them. That is not an unheard-of fault.
You might even change all seven of those components just in case it has been blasted one way or another at some point. I would use a 35V or higher capacitor for C41 if you do. 1 volt of headroom is pushing it a bit.
Also be sure to ensure that the board is totally clean everywhere. Those things have a nasty habit of doing all sorts of strange things like growing corrosive, conductive fuzzies and such from exposure to chlorine vapours.
I'm wondering whether D25 is really necessary. The application circuit in the datasheet has no regulation or external overvoltage protection on the Vcc pin (although there is an internal 16V zener clamp). In any case, Q9 and D26 circumvent any potential OV problem. Moreover, the run current for the IC is only 2mA, and almost all of that is supplied via R46 and R47.
Sorry I went missing for Christmas, but had lots of food to eat and beers to drink... Hope you all enjoyed your Christmas too!
Lots of food for thought here, thanks gents. I think there is a mistake in the Fairchild reference schematic where the cathode of 16V zener CR9 is meant to connect to VCC pin 7 and not I limit pin 6 as drawn, meaning the Zodiac circuit is more similar to reference than first thought, except for 24V zener instead of 16V zener, 10K R21 in shunt with zener and a 10R R31 current limiter from inductor winding in reference circuit.
In another incredible coincidence, while my Zodiac Tri Chlorinator is out of action, I am using a friends borrowed spare Zodiac Tri that had exactly the same zener fail in it not long after mine failed originally.
I just happened to check it while it was running today and it seems to have failed again.

So for some reason, both these units operated for 7 -8 years with no trouble, and suddenly this zener failure seems to be a regular fault, which surely must be caused by the decline or fault of another component. I will be removing the components suggested and testing/replacing.